Cover page

The Hubble Space Telescope has brought us fabulous images of our universe. Although Galaxies, like our

Milkyway Galaxy, appear very quiet, they are not. We are falling into our nearest Andromeda Galaxy, and on

one ﬁne day, it will result in a vast collision when the two will rip oﬀ parts and merge to form a single elliptical

galaxy!

Such galaxy mergers are not rare. The picture of Antennae galaxies on the cover page is such a merger as

captured by Hubble. It is about 45 million light years away from us, and it took about a hundred million years

for them to fall under gravity into each other! (image courtesy Hubble Space Telescope)

Meger does not mean the end of a galaxy. What you see as bright pink and blue spots in the image are newborn

stars formed due to the strong shockwaves created by the merger. There are billions of them! The orange blobs

are parts of the original galaxies, and the brown regions are composed of dust in them.

Managing Editor Chief Editor Editorial Board Correspondence

Ninan Sajeeth Philip Abraham Mulamootil K Babu Joseph The Chief Editor

Ajit K Kembhavi airis4D

Geetha Paul Thelliyoor - 689544

Arun Kumar Aniyan India

Jorunal Publisher Details

Publisher : airis4D, Thelliyoor 689544, India

Website : www.airis4d.com

Email : nsp@airis4d.com

Phone : +919497552476

Editorial

by Fr Dr Abraham Mulamoottil

airis4D, Vol.1, No.3, 2023

www.airis4d.com

India Aims an Entrepreneurship-ready Education

Future Education in India, as per the National Education Policy (NEP) 2020, provides importance for

entrepreneurship and start-ups. The traditional approach towards education was accused of creating students

to acquire degrees, government jobs (“sepoy and Macaulayism”) and preparing manual workers for industrial

needs. However, NEP aims for a knowledge revolution where atleast 50% of the society should be made

knowledge workers by 2040. This third edition of airis4D journal discusses research based innovations and

entrepreneurship-ready products which can set up enterprises and start-ups. The article “Generative Act“ is an

exciting application of Machine Learning models with huge artistic and commercial value. The Nuts and Bolts

of Machine Learning has become relevant in several industries like healthcare, ﬁnance and education. Investing

time and eﬀort in obtaining a solid knowledge of these principles and approaches is a worthy endeavour for

anybody interested in Machine Learning and may result in substantial ﬁeld improvements. “The art of Data

Transformation” guides transforming text into features. Sky was considered an abode of Gods and Goddesses

in the beginning of our civilisation. But now it is exciting, and it gives the fundamental knowledge of human

life and life of the planets. So, humanity cannot limit astronomic researcher till the end of life. The frontiers of

astronomy research opened ﬂoodgates of data and its applications that give new knowledge about our life and the

life of the universe. It is studied that stars are luminous spheroids of plasma held together by its gravity. If a star’s

brightness, as seen from Earth, changes with time, it is called a Variable Star. In biology, CRISPER Cas9 – the

gene therapy now applied in neurogenerative disorders, genetic diseases, diagnostics and therapeutics in cancer,

blindness etc. LiDAR Imaging has multiple applications in daily life like Forestry, Navigation, Archaeology

and Hydrology. Finally, the ﬁction “Curves and Curved Trajectories” is a thought experiment on a virtual world

with imaginary creatures capable of doing ﬁctitious activities.

In this context, discussion on future education fundamentally focuses on entrepreneurship and start-ups.

Academy has to discuss not the issues of job seekers but the job creators. Young people are looking to set startups,

and they are keen and passionate about being founders and cofounders. Entrepreneurship-ready education refers

to an approach to education that prepares students to become successful entrepreneurs. This education focuses

on developing students’ skills, knowledge, and mindset to create and grow a successful business.

An entrepreneurship-ready education typically includes courses in business management, marketing, ﬁ-

nance, and accounting, as well as courses in innovation, creativity, and problem-solving. It also involves

experiential learning opportunities, such as internships, mentorship programs, and hands-on projects, that allow

students to apply what they have learned in real-world situations. An essential aspect of entrepreneurship-ready

education is the development of an entrepreneurial mindset. This mindset involves a willingness to take risks,

an ability to identify and pursue opportunities, a desire to learn from failure, and a commitment to innovation

and continuous improvement.

Entrepreneurship-ready education can be beneﬁcial not only for those who plan to start their businesses

but also for those who wish to work in established companies. The skills and mindset developed through

entrepreneurship-ready education can help individuals to be more innovative, adaptable, and eﬀective in any

professional setting.

Finally, the future of education in India will likely be shaped by several factors, including technological

advancements, changing social and economic conditions, and evolving pedagogical practices. Besides, future

teachers shall work like starters’ mentors to guide and nurture the students, fostering their creativity, innovation,

and critical thinking skills. India in this century is vibrant, having a dynamic and young population, with 65%

of Indians being under 35 years old. The articles in this third edition of airis4D journal may provide ideas for

students to initiate entrepreneurship-ready startups and be founders and cofounders.

iii

Contents

Editorial ii

I Artiﬁcial Intelligence and Machine Learning 1

1 Generative Art 2

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Generative images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Generative music . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.4 Generative text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Nuts and Bolts of Machine Learning - Part 1 10

2.1 Breaking into the black box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.2 Beyond curve ﬁtting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3 The Building Blocks of Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Text to Features: The Art of Data Transformation 15

3.1 Bag of Words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 TF - IDF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.3 Word2vec . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

II Astronomy and Astrophysics 21

1 Why the Sky is not the Limit for Astronomy Research? 22

1.1 What is Astronomy? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.2 Astronomy in the Ancient Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1.3 The Age of Scientiﬁc Revolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

1.4 Astronomy in The Space Race Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

1.5 The Impact of Astronomy Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

1.6 The story behind the GPS and the Wireless Internet . . . . . . . . . . . . . . . . . . . . . . . 29

1.7 Astronomy in the Age of Big Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

1.8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2 Variable Stars 31

2.1 What are variable stars? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.2 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.3 Classiﬁcation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

CONTENTS

III Biosciences 35

1 CRISPER Cas9 - The Gene Therapy 36

1.1 What is the CRISPR-Cas9 System ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1.2 The Editing Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

1.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

IV General 40

1 LiDAR Imaging 41

1.1 LiDAR system components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

1.2 Measuring trees with light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

1.3 Full Waveform vs Discrete LiDAR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

1.4 Applications of LiDAR Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

1.5 Advantages and Disadvantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

V Fiction 46

1 Curves and Curved Trajectories 47

Part I

Artiﬁcial Intelligence and Machine Learning

Generative Art

by Arun K Aniyan

airis4D, Vol.1, No.3, 2023

www.airis4d.com

1.1 Introduction

Machine learning is deﬁned as teaching computers to imitate human wisdom, which means gaining the

ability to judge and make decisions. This had been the whole crux of machine learning technology for several

decades until 2014 when a young Google intern Alexander Mordvintsev presented a breakthrough result.

This was a period of proliferation of deep convolutional neural networks and major breakthroughs were

made in the area of machine learning and speciﬁcally deep learning. Alexander was experimenting with a deep

learning model called Inception. He trained the model with a large amount of image data for classiﬁcation

purposes. In other words, the model was trained to predict a class label for a given input image. For example,

if the input image was that of a ﬂower, the model should output the class label ”ﬂower”. A sample model is

shown in Figure 1

Figure 1: A representation of a simple convolution neural network which takes an input image and predicts

known 10 class labels at the output.

In such models, information ﬂows from left to right. Alexander took such a network and tweaked the model

to understand what the model perceives as features for a given class of objects. In actual practice, this procedure

is more complex than mentioned here, but for simplicity let us imagine the model was tweaked to rearrange

random noise input to adjust the noise pixels to closely resemble what it would think as a speciﬁc class. The key

1.2 Generative images

idea is to tweak the diﬀerent layers of the model so that it can produce an image from the intermediate layers.

So for example, a noise pixel image would be given as input and the model parameters would be adjusted to

produce what it would think is a dog. The results resembled psychedelic images. This method is known as

“Deep Dream” mainly because it was similar to allowing a deep learning model to dream. Figure 2 shows an

example of a Deep dreamt image.

Figure 2: An example of a deep dream of a dog image. This generated image is also called “dogception”

because it is generated by a model named “inception”. [Image credit: Google Blogs]

This result with Inception networks paved a new stream of applications with deep learning networks.

DeepDream was in fact the ﬁrst demonstration of generating artistic “image” work generated by a machine

learning model even though researchers had this idea for several years.

1.2 Generative images

A major leap of progress has been made in the area of generative images since DeepDream was invented.

Soon after DeepDream models, Ian Goodfellow developed a seminal machine learning model called Generative

Adversarial Network (GAN). The initial intent of this model design was to generate realistic fake data points to

improve the generalization ability of deep neural networks. A GAN model has two primary components which

are “Generator” and “Discriminator”. The discriminator tries to identify the fake data points generated by the

generator. The generator’s job is to produce near-realistic data to fool the discriminator. Machine learning

researchers used this ability of a generator model to generate fake images of objects. One of the ﬁrst attempts

was to generate fake human faces or in other words person who does not exist. Since 2014, the progress in

generating fake human faces has been extremely impressive. The diﬀerent generations of human face generation

are shown in Figure 3.

The latest human face-generating algorithms generate extremely powerful in the sense that it has become

challenging to identify if they are artiﬁcially generated. There are even websites that generate random faces of

1.3 Generative music

Figure 3: ’The diﬀerent generations of human face’ generated using GAN models.

people who do not exist.

Digital art (also referred to as electronic art) is largely generated either by vector graphics, ray tracing, and

other graphical tools. Very recently with the advent of machine learning models called “Diﬀusion models”,

digital art in the form of images has jumped to a completely diﬀerent scale and level. Diﬀusion models take a

textual description called “prompt” as input and generates an image based on the text description as the output.

These are complex models that are trained on a very large set of images with text descriptions. Based on the

text data per image, the model has learned objects’ diﬀerent patterns, styles, and relations. For example, Figure

4 shows the generated images by a diﬀusion model for the input text “River ﬂowing in from a galaxy center

ﬁlled with unicorns and fairies dancing around to a mystical beautiful world in the ultra-high deﬁnition”.

As shown in the example, the details in the image are so much realistic that the colours, placement, and

shape of objects match very well with an image generated by a human. The machine learning model has picked

up all such details during the training phase when an input prompt is given, it will randomly generate the images

keeping the context of the image appropriate. The example image is a “cartoon” style image. But there are

advanced diﬀusion models which can generate ultra-realistic images which resemble those that are directly

taken with a professional camera. It is important to note that the realistic quality of the result images not only

depends on the model but also on the text prompt given. As of now “prompt engineering” is considered as an

artistic career path, because people are paid to generate text prompts to produce hyper-realistic AI-generated

art. Figure 5 shows some extremely realistic machine-generated images.

As of now, there are a plethora of online tools available for free as well as paid to generate artiﬁcial images.

Midjourney and DreamBooth are two of the most popular tools. With these tools, people are not just generating

images, but multiple frames to generate short and long videos which would otherwise cost a lot. Social media

is now ﬂooded with such images and videos which has generated a new class of visual artists.

1.3 Generative music

Images are not the only area where machine learning models are used to generate art, music is as well

generated in a similar manner. Machine learning models for music generation are trained with music from

diﬀerent genres, artists, and regional styles. This has enabled people to generate music that is royalty-free which

a lot of artists around the world use for background scores, promotions, and diﬀerent purposes.

Training models for music and speech processing have been a long ongoing research. But still, there were

challenges in generating music that was very natural in terms of tones and mood. With the breakthrough in deep

learning, this area just like others made a huge step in producing very tangible results With widely available

music data in digital form, data is available in plenty to train such large complex models.

There are websites such as juke box, sound draw and aiva which oﬀers service to generate your own music

1.3 Generative music

Figure 4: Sample images generated by a diﬀusion model for given input text prompt.

1.3 Generative music

Figure 5: Some hyper-realistic images created by a diﬀusion model based on text prompts.

1.4 Generative text

by providing details like genre, artist, and style.

One of the most compelling factors of such tools is that in addition to the actual music, the vocal with lyrics

can also be generated. The generated music will resemble a person singing with proper lyrics and a background

score. If you visit one of the earlier mentioned websites you will be able to play and hear the best sample

music generated by machine learning models. One of the latest machine learning models published by Google

research is called the MusicLM which can generate music from text and is able to generate high-ﬁdelity music

scores.

1.4 Generative text

Natural language processing is considered one of the most challenging areas in artiﬁcial intelligence

research. All those expressions, thoughts, and ideas generated in the human brain are expressed best through

language. Training machine learning models to replicate understanding of language is a complex process.

Natural language comprises three major components namely, syntax, semantics, and context. It is only when

these three components form a speciﬁc pattern a comprehensible expression can be made with language. And

most importantly these patterns vary across diﬀerent languages and dialects.

Sequence models based on Long-Short Term Memory (LSTM) are generally employed for language

modeling. In 2016, a seminal work introduced the concept of “attention networks” which gave rise to a class

of models called “transformer” models. Transformer models completely changed the way natural language

processing was done. Recently there are a class of models called “Large Language Models (LLM)” that are

able to represent language understanding close to human levels.

LLMs are a hot topic of research and discussion in the machine-learning world and are used to generate

synthetic text based on prompt inputs. OpenAI foundation is the pioneer in LLMs and they released models

such as Generative Pre-trained Transformer (GPT) and ChatGPT which are able to generate text just like a

human. These models were trained on an extremely large corpus of text data originating from various sources

including source code from diﬀerent programming languages. Such a model will have learned both the content

and context from the diﬀerent sources and when you try to generate some output, it will resemble a well-read

person replying to you. ChatGPT is a recent LLM that has shown incredible results in terms of prompt response

and text generation. As the name implies, ChatGPT responds like a chatbot. Shown in Figure 6 is an example

where ChatGPT responds to a code snippet.

In the shown example ChatGPT is able to generate such a response because, during the training phase, it

has been exposed to corpus data from programming textbooks as well as general English. This simply means

that ChatGPT is exposed to diﬀerent writing and narration styles and also has some level of context with the

previous sequence of text.

In the context of generative text, models like ChatGPT and other LLMs have been used to automatically

generate stories and other literature for very speciﬁc artworks. For example, in Figure 7, I asked ChatGPT to

write the ﬁrst paragraph of Othello in Indian Mythological style.

These sort of LLMs has an immense amount of possibilities in terms of generating creative text. One

important point to note about such models is that they are not perfect and create plenty of mistakes. The biases

in the data will also aﬀect the model output.

1.5 Conclusion

Figure 6: ChatGPT responding to a code snippet with a bug.[Image Credits: Twitter]

Figure 7: ChatGPT writing the ﬁrst paragraph of Othello in Indian mythological style.

1.5 Conclusion

Generative art is an exciting application of machine learning models which has huge artistic as well as

commercial value. These models even though in the early stages are producing impressive results, still require

substantial research to reduce the data biases and also inconsistencies in the results. For example for certain

generative images with human subjects in them, some models tend to generative additional ﬁngers on limbs.

One general concern with such models is the amount of misuse that could happen. The majority of these

models are publicly available and people could use them for unethical purposes. Apart from the public API’s,

the publicly available ones have no locks for profanity checks which is another dangerous situation.

But to conclude, generative art has given rise to a new set of artistic and creative tools which helps create

royalty-free content.

References

Inceptionism: Going Deeper into Neural Networks

Goodfellow, Ian J., Jonathon Shlens, and Christian Szegedy. ”Explaining and harnessing adversarial

examples.”

Cheng, Keyang & Tahir, Rabia & Eric, Lubamba & Li, Maozhen, An analysis of generative adversarial

networks and variants for image synthesis on MNIST dataset. Multimedia Tools and Applications, 2020

Peebles, William, and Saining Xie. ”Scalable Diﬀusion Models with Transformers.”

Briot, Jean-Pierre, Ga

etan Hadjeres, and Franc¸ois-David Pachet. ”Deep learning techniques for music

generation–a survey.”

Gers, Felix A., J¨urgen Schmidhuber, and Fred Cummins. ”Learning to forget: Continual prediction with

LSTM.” Neural computation 12.10 (2000): 2451-2471

Brants, Thorsten, et al. ”Large language models in machine translation.” (2007)

Floridi, Luciano, and Massimo Chiriatti. ”GPT-3: Its nature, scope, limits, and consequences.” Minds

and Machines 30 (2020): 681-694.

Budzianowski, Pawe l, and Ivan Vulic. ”Hello, it’s GPT-2–how can I help you? towards the use of

pretrained language models for task-oriented dialogue systems.”

About the Author

Dr.Arun Aniyan is leading the R&D for Artiﬁcial intelligence at DeepAlert Ltd,UK. He comes from

an academic background and has experience in designing machine learning products for diﬀerent domains. His

major interest is knowledge representation and computer vision.

Nuts and Bolts of Machine Learning - Part 1

by Blesson George

airis4D, Vol.1, No.3, 2023

www.airis4d.com

2.1 Breaking into the black box

In recent years, there has been a signiﬁcant increase in the use of machine learning (ML) techniques

across a wide range of ﬁelds, from ﬁnance and healthcare to marketing and customer service. While ML has

the potential to provide powerful insights and improve decision-making, there is a growing concern that many

people are using ML as a black box, without fully understanding how the models arrive at their predictions or

decisions.

The term ”black box” refers to the idea that the model is like a sealed box where you can observe the inputs

and outputs, but you don’t have access to the internal workings of the system. Sometimes, the decision-making

process of a machine-learning model is referred to as a black box since the researchers and users generally know

the inputs and outputs, but it is diﬃcult to observe what is happening within. This is because, in conventional

programming, users can frequently follow the code and understand how the program is making decisions or

creating results. With machine learning, however, the model is trained on data and generates its own internal

representations and patterns, which can be challenging for users to interpret and comprehend.

When ML is seen as a black box model, there are several diﬃculties. The lack of transparency is the major

drawback of the black box technique. It becomes diﬃcult to verify that an algorithm is making judgements in

an ethical, fair, and consistent manner with our beliefs when we cannot comprehend how it is making these

decisions. Machine learning algorithms can inherit bias from the data on which they were trained, and it can be

diﬃcult to discover and adjust for this bias if the method is opaque. This can have discriminatory consequences,

particularly for underprivileged populations. The black box method might make it diﬃcult to explain how a

choice or forecast was reached, particularly when the conclusion has substantial ramiﬁcations. This can make

it tough to discover problems and make algorithm improvements. The black box approach can also make it

diﬃcult to recognise and avoid assaults on the algorithm, such as adversarial attacks in which an adversary feeds

the algorithm intentionally misleading or incorrect data to alter its decisions.

Understanding machine learning beyond the black box involves gaining a deeper understanding of the

internal workings of machine learning models, including their structure, assumptions, and limitations. Here are

some steps you can take to gain a deeper understanding of machine learning:

1. Learn the basics: Start by building a strong foundation in the basic concepts and principles of machine

learning. This includes understanding the diﬀerence between supervised and unsupervised learning, the

importance of training and testing data, and the diﬀerent types of machine learning models.

2. Understand the math: Machine learning is built on a foundation of mathematical concepts, including

2.2 Beyond curve ﬁtting.

linear algebra, calculus, and probability theory. Gaining a deeper understanding of these concepts can

help you understand the inner workings of machine learning models and how they make decisions.

3. Analyze the data: Machine learning models are only as good as the data they are trained on. Analyzing the

data can help you understand its quality, identify patterns and relationships, and make informed decisions

about feature selection and engineering.

4. Interpret the results: Instead of treating the output of a machine learning model as a black box, try to

interpret the results in the context of the problem you are trying to solve. This can help you identify areas

where the model is performing well, as well as areas where it may be making errors.

5. Visualize the model: Many machine learning models can be visualized in a way that makes it easier to

understand how they are making decisions. This can include visualizing decision boundaries, feature

importances, or the structure of a neural network.

6. Experiment and evaluate: Machine learning is an iterative process, and experimenting with diﬀerent

models, hyperparameters, and features can help you gain a deeper understanding of how machine learning

works. Evaluating the performance of diﬀerent models on a test set can also help you identify areas where

the model can be improved.

By taking these steps, one can move beyond the black box and gain a deeper understanding of the inner workings

of machine learning models. This can help in building more eﬀective and robust models, as well as make more

informed decisions about how to use machine learning in work.

2.2 Beyond curve ﬁtting.

Machine Learning (ML) is a subﬁeld of Artiﬁcial Intelligence (AI) that focuses on developing algorithms

and models that can learn and make predictions or decisions based on data. It has gained signiﬁcant attention

and popularity in recent years due to its ability to solve complex problems and make accurate predictions.

However, there is a common misconception among some people that ML is nothing more than gloriﬁed curve

ﬁtting, which can lead to misunderstandings and underestimation of its capabilities. This misconception arises

from the fact that ML algorithms often involve ﬁtting a model to data, similar to curve ﬁtting techniques used in

mathematics and statistics. In this context, the model is often seen as a mathematical function that approximates

the underlying relationship between input variables and output variables. However, ML is much more than just

curve ﬁtting, as it involves complex algorithms and techniques such as deep learning, reinforcement learning, and

unsupervised learning, which can handle high-dimensional data and learn complex patterns and relationships.

This essay will delve deeper into the misconceptions surrounding ML as gloriﬁed curve ﬁtting and explain why

it is much more than that.

Curve ﬁtting is a technique used to model a relationship between a set of input features and corresponding

output values, by ﬁnding the best function or curve that ﬁts the data. The goal is to ﬁnd a function that can

accurately predict the output value for a given input value. In curve ﬁtting, the data is often represented as a set

of points in a coordinate system. The goal is to ﬁnd a curve that passes as close as possible to these points.

In the case of ﬁtting a straight line to two points, the model would be a linear equation of the form y = mx

+ b, where m is the slope of the line and b is the y-intercept. To ﬁt a straight line to two points, we need to

determine the values of m and b that will make the line pass through both points.

Curve ﬁtting can be a powerful tool for making predictions, estimating parameters, and analyzing data.

However, there are several limitations to curve ﬁtting that should be considered:

1. Overﬁtting: Curve ﬁtting can result in overﬁtting, where the model ﬁts the training data too closely and

fails to generalize well to new data. This can lead to a loss of predictive power and inaccurate results.

2.3 The Building Blocks of Machine Learning

2. Underﬁtting: Curve ﬁtting can also result in underﬁtting, where the model is too simple to capture the

underlying patterns in the data. This can lead to poor performance and inaccurate results.

3. Assumptions: Curve ﬁtting is based on certain assumptions about the data, such as linearity, normality,

and independence. If these assumptions are violated, the results of curve ﬁtting can be biased or incorrect.

4. Extrapolation: Curve ﬁtting is often used to make predictions beyond the range of the observed data.

However, extrapolation can be risky and may result in inaccurate or unreliable predictions.

5. Data quality: Curve ﬁtting is only as good as the data that is used to ﬁt the model. If the data is incomplete,

inconsistent, or contains errors, the results of curve ﬁtting may be unreliable.

Machine learning (ML) is indeed much more than curve ﬁtting. While curve ﬁtting is a mathematical

technique used to ﬁnd a function that best ﬁts a set of data points, machine learning is an approach to building

computational models that can learn and make predictions or decisions based on data. Machine learning models

are typically used in cases where the relationship between the input data and the output is too complex or

unknown to be captured by a simple mathematical function, such as in cases where there are many input features

or non-linear relationships.

Machine learning includes a range of techniques and models, such as linear regression, decision trees,

random forests, support vector machines, neural networks, and deep learning. These models diﬀer in their

structure and complexity, but they all share the common goal of learning from data and making predictions or

decisions based on that data.

In addition to predictive modeling, machine learning is also used in other areas such as clustering, anomaly

detection, reinforcement learning, natural language processing, and computer vision. These applications

of machine learning involve tasks such as grouping similar data points, identifying unusual patterns or events,

teaching agents to perform tasks through trial and error, processing and understanding language, and recognizing

objects in images.

Machine learning is a powerful and versatile approach to building computational models that can learn and

make predictions or decisions based on data. While curve ﬁtting is a useful mathematical tool, it is just one of

many techniques used in machine learning and cannot capture the full scope and complexity of this ﬁeld.

2.3 The Building Blocks of Machine Learning

Machine Learning involves the development of algorithms and models that can learn and make predictions

based on data. However, understanding the underlying nuts and bolts of ML is crucial for researchers and

practitioners to be able to eﬀectively develop and apply these models. In this context, the ”nuts and bolts” refer

to the fundamental concepts, techniques, and algorithms that form the building blocks of machine learning.

These include data preprocessing, feature selection, model selection, hyperparameter tuning, and performance

evaluation. Each of these components plays a critical role in the development and application of ML models,

and mastering them is essential for anyone interested in machine learning.

2.3.1 Data preprocessing

Data preprocessing is a crucial step in machine learning (ML) that involves cleaning, transforming, and

preparing raw data before it can be used to train an ML model. Raw data is often messy, incomplete, and

inconsistent, which can adversely aﬀect the performance and accuracy of ML models. Data preprocessing aims

to address these issues by transforming the raw data into a format that is suitable for analysis and modeling.

The process of data preprocessing typically involves several steps, including data cleaning, data transfor-

2.3 The Building Blocks of Machine Learning

mation, and data reduction. Data cleaning involves removing or correcting errors, ﬁlling in missing values, and

removing outliers. Data transformation involves converting data into a suitable format, such as scaling numeric

features or encoding categorical features. Data reduction involves selecting a subset of relevant features or

reducing the dimensionality of the data to improve model performance and reduce computational complexity.

Data preprocessing is essential for building accurate and robust ML models. Poor quality data can result

in inaccurate predictions and models that are not generalizable to new data.

2.3.2 Feature selection

Feature selection is a critical step in machine learning (ML) that involves selecting the most relevant and

informative features from a dataset. In many cases, datasets contain a large number of features that may not all

be relevant or useful for training an ML model. Feature selection aims to reduce the dimensionality of the data

by selecting a subset of features that are most important for predicting the target variable.

There are several techniques for feature selection, including ﬁlter methods, wrapper methods, and embedded

methods. Filter methods involve ranking features based on statistical measures such as correlation or mutual

information and selecting the top-ranked features. Wrapper methods involve selecting features based on the

performance of an ML model trained on diﬀerent subsets of features. Embedded methods involve selecting

features during the training process of an ML model, where the importance of each feature is learned and used

to guide the training process.

Feature selection is important for several reasons. First, it can improve the performance of ML models

by reducing the noise and redundancy in the data. Second, it can speed up the training process and reduce

computational complexity by reducing the dimensionality of the data. Third, it can improve the interpretability

and explainability of ML models by focusing on the most important features.

However, it’s important to note that feature selection is not always necessary or beneﬁcial for every dataset

or ML problem. In some cases, keeping all features may be necessary for achieving optimal performance.

Therefore, it’s essential to carefully evaluate and experiment with diﬀerent feature selection techniques to

determine the best approach for a particular ML problem.

2.3.3 Model Selection

Model selection is an important step in the process of building a machine learning model. The goal of

model selection is to choose the best model out of a set of candidate models for a given task, based on certain

performance metrics.

Model selection is a crucial step in machine learning (ML) that involves choosing the best algorithm or

model to use for a given problem. There are many diﬀerent ML algorithms and models to choose from, each

with their strengths and weaknesses, and selecting the best one can have a signiﬁcant impact on the performance

and accuracy of the ﬁnal model.

The process of model selection typically involves several steps, including deﬁning the problem, selecting

a set of candidate models, evaluating the models using appropriate metrics, and selecting the best-performing

model. Some common techniques for model selection include cross-validation, grid search, and randomized

search.

Cross-validation involves dividing the dataset into several subsets and using each subset as a test set while

training the model on the remaining subsets. This allows for the evaluation of each model’s performance on

multiple subsets of the data and helps to reduce the risk of overﬁtting.

2.4 Conclusion

Grid search involves systematically evaluating diﬀerent combinations of hyperparameters for each model

and selecting the combination that results in the best performance. Randomized search is similar to grid search

but involves randomly selecting combinations of hyperparameters to evaluate, which can be more eﬃcient for

larger datasets.

2.4 Conclusion

Understanding the core principles and techniques of machine learning (ML) is essential for both academics

and practitioners. From data preparation through feature selection, model selection, hyperparameter tweaking,

and performance evaluation, each of these components is essential to the creation and use of machine learning

(ML) models. These building components must be mastered in order to construct ML models that are accurate,

resilient, and capable of handling real-world issues.

Mastering the fundamentals of ML has become more important, given the growing relevance of machine

learning in several industries, like healthcare, ﬁnance, and education, among others. So, investing time and

eﬀort in obtaining a solid knowledge of these principles and approaches is a worthy endeavour for anybody

interested in ML, and may result in substantial ﬁeld improvements.

References

Breaking into the black box of artiﬁcial intelligence

Is Machine Learning just gloriﬁed curve ﬁtting?

The Scope and Limitations of Curve Fitting

Feature Engineering for Machine Learning: Principles and Techniques for Data Scientists

About the Author

Blesson George is currently working as Assistant Professor of Physics at CMS College Kottayam,

Kerala. His research interests include developing machine learning algorithms and application of machine

learning techniques in protein studies.

Text to Features: The Art of Data

Transformation

by Jinsu Ann Mathew

airis4D, Vol.1, No.3, 2023

www.airis4d.com

How to Convert Text into Features for Machine Learning Models? As an NLP(Natural Language Process-

ing) beginner, have you ever faced the challenge of turning text entries into features for a Machine Learning

model? This can be a common question when working with text data for the ﬁrst time. In this article, we will

provide an answer to this question and guide you through the process of transforming text into features.

Features in Machine Learning are simply numerical values that allow mathematical operations such as

matrix factorization, dot products etc. In Natural Language Processing, Feature Extraction is a crucial step in

gaining a deeper understanding of the text data. After cleaning and normalizing the initial text, it is important

to transform it into features for use in modeling, as the machine does not compute textual data. So, we go for

numerical representation of individual words for the ease of processing by computers.This transformation task

is generally called feature extraction of document data. There are various ways to perform feature extraction,

some popular and mostly used are:-

Bag of words

Tf Idf

Word2vec

3.1 Bag of Words

The bag of words approach is a straightforward technique for extracting features from text that can be

utilized in machine learning algorithms. The basic idea is to create a bag of words, or corpus, by collecting

all the words in the text. In this approach, each text document is represented as a numeric vector, with each

dimension corresponding to a particular word from the corpus. The value of each dimension may reﬂect the

frequency of the word in the document, its occurrence (indicated by 1 or 0), or a weighted value. The name of

this model is derived from the fact that each document is represented as a ’bag’ of its own words, with no regard

for word order, sequence, or grammar. The idea behind this method is simple: to represent sentences as vectors

by counting the frequency of the words they contain

Here’s an example of how to create a bag of words from a sentence:

Sentence : ”The cat sat on the mat”

⋄ Tokenize the sentence by breaking it into individual words: ”The”, ”cat”, ” ”sat”, ”on”, ”the”, ”mat.”

⋄ Create a list of unique words in the sentence, in alphabetical order: ”cat”, ”mat”, ”on”, ”sat”, ”the”.

3.2 TF - IDF

Figure 1: Bag of words representation of the sentence ‘The cat sat on the mat’

⋄ Count the frequency of each word in the sentence: ”cat” (1), ”mat” (1), ”on” (1), ”sat” (1), ”the” (2).

⋄ Represent the sentence as a bag of words vector, with each dimension corresponding to a word in the list

and the value reﬂecting its frequency: [1, 1, 1, 1, 2].

“The cat sat on the mat” = [1,1,1,1,2] This should make things clearer. The bag of words representation

for the sentence ”The cat sat on the mat” would be [1, 1, 1, 1, 2], with the dimension for ”cat” having a value of

1, the dimension for ”mat” having a value of 1, and so on (Figure : 1).

Drawbacks of Bag Of Words

Even Though Bag-of-Words method is a popular approach for representing text data in natural language

processing (NLP) tasks, it has several drawbacks, including:

• Ignoring word order: The BoW method treats each word as a separate feature and ignores their order.

This can result in loss of important information, such as the context and meaning of words within a

sentence or document.

• Vocabulary size: The BoW method generates a large number of features, which can result in a high-

dimensional feature space. This can make it diﬃcult to process and analyze the data eﬃciently, especially

when dealing with large datasets.

• Sparsity: In the BoW method, most of the features have zero values because most words occur only a

few times in a document or corpus. This results in a sparse representation, which can aﬀect the accuracy

of models that use the BoW method.

• Lack of semantic understanding: The BoW method does not capture the meaning or semantics of

words, and treats all words as independent entities. This can limit the accuracy and usefulness of NLP

models that rely solely on the BoW method.

• Inability to handle out-of-vocabulary words: The BoW method relies on a ﬁxed vocabulary, which

means that any words that are not included in the vocabulary are ignored or treated as unknown. This can

result in loss of information and aﬀect the accuracy of NLP models.

Overall, the Bag-of-Words method can be a useful and eﬀective approach for text data representation in NLP

tasks, but it is important to be aware of its limitations and to consider alternative methods that can address these

drawbacks.

3.2 TF - IDF

(image courtesy:https://betterprogramming.pub/a-friendly-guide-to-nlp-tf-idf-with-python-example-5fcb26286a33

Figure 2: Term frequency and inverse document frequency calculation

3.2 TF - IDF

TF-IDF stands for ”term frequency-inverse document frequency.” It is a numerical statistic used in infor-

mation retrieval and natural language processing to evaluate the importance of a word in a document, relative

to its occurrence in other documents.

TF-IDF is based on two main factors:

• Term Frequency (TF): Term frequency is a measure of how frequently a term (word) occurs in a

document. It can be computed by simply counting the number of occurrences of the term within the

document. However, since the calculation of term frequency is aﬀected by document length and the

generality of the term, some adjustments are necessary to avoid biased results. This is because the same

term can appear more frequently in a longer document than in a shorter one. To overcome this problem, the

raw count of the term in the document is typically divided by the document length to obtain a normalized

term frequency value.

Term Frequency = No:of times a term(word) appeared in the document/(length of the document)

• Inverse Document Frequency (IDF): This measure tells how rare (or common) a term is in the entire

set of documents. While computing term frequency, all terms are treated equally, which can lead to

misleading results because frequently occurring but unimportant words (such as ”is,” ”of,” and ”that”)

may overshadow rarer but more signiﬁcant words. To address this issue, frequent terms are scaled down

and rare terms are scaled up by applying inverse document frequency. This measure is a logarithmic

transformation of a fraction, obtained by dividing the total number of documents in the corpus by the

number of documents that contain the word of interest.

Inverse Document Frequency = log[No:of documents in the corpus /(No:of documents with a particular

word in it)]

The TF-IDF score is the product of the term frequency and inverse document frequency (Figure:2). This score

gives a measure of the importance of a word in a speciﬁc document, with higher scores indicating greater

importance. It is often used as a feature in machine learning algorithms for text classiﬁcation and information

retrieval systems.

Example of TF-IDF calculation:

Let’s say we have a set of three documents:

Document 1: ”The cat in the hat”

Document 2: ”The mouse in the house”

Document 3: ”The dog in the fog”

3.3 Word2vec

We want to calculate the TF-IDF score for the term ”cat” in Document 1.

First, we calculate the term frequency (TF) for ”cat” in Document 1. The term ”cat” appears only once in

Document 1, so the raw term frequency is 1.

Next, we calculate the inverse document frequency (IDF) for ”cat” across all documents. In our example,

the term ”cat” appears in only one document (Document 1), so the IDF is:

IDF(”cat”) = log (Total number of documents / Number of documents containing ”cat”)

= log (3 / 1)

= log(3)

= 1.09

Finally, we calculate the TF-IDF score for ”cat” in Document 1 by multiplying the term frequency and inverse

document frequency:

TF-IDF(”cat”, Document 1) = TF(”cat”, Document 1) x IDF(”cat”)

= 1 x 1.09

= 1.09

Therefore, the TF-IDF score for the term ”cat” in Document 1 is 1.09.

Drawbacks of TF-IDF

• Limited semantic understanding: TF-IDF only considers the frequency of individual terms, and does

not capture the meaning of the text. It treats each word as an independent entity, without considering

the context in which it appears. This can lead to inaccurate results when dealing with highly ambiguous

words or phrases.

• Computationally expensive: The calculation of TF-IDF can be computationally expensive, especially for

large collections of documents. This can be a challenge in applications that require real-time processing

or near-instantaneous response times.

• Vocabulary mismatch: The eﬀectiveness of TF-IDF can be limited by a vocabulary mismatch between

the query and the document. If the words used in the query do not match the words in the document, the

TF-IDF score may not accurately reﬂect the relevance of the document.

3.3 Word2vec

Although Bag-of-Words and TF-IDF have been widely used in natural language processing, they do not

account for the contextual meaning of words and can become overly complex when dealing with a large

vocabulary. To address these limitations, a more eﬃcient method is needed that can control the size of word

representations and capture semantic relationships between words. This is where word2vec comes in.

Word2vec model was developed by Google in 2013 and it generates dense, high-quality vector representa-

tions of words. These embeddings capture both the semantic and contextual similarity of words and are created

using large-scale textual corpora. Word2vec is an unsupervised model that can construct a vocabulary of words

and generate a dense word embedding for each word in the vocabulary.

Word2vec operates as a two-layer neural network that processes text as words. The input to the network is

the text corpus, and the output is a set of feature vectors that represent the words in the corpus. The generated

vector representations are of a ﬁxed size, and Word2vec arranges them in a way such that similar words are

closer together, and opposite words have the same diﬀerence in distance, if applicable (Figure: 3)

3.3 Word2vec

(image courtesy:https://www.analyticsvidhya.com/blog/2021/07/word2vec-for-word-embeddings-a-beginners-guide/

Figure 3: Similar words are closely placed in vector space

CBOW and Skip-gram are two methods used in the Word2Vec algorithm, which is a way of representing

words in a numerical form that computers can understand.

CBOW predicts a target word based on the words around it, while Skip-gram predicts the words around a

target word (Figure: 4) For example, let’s say we have the sentence ”The cat sat on the mat”. In CBOW, the

model would be trained to predict the word ”sat” based on the surrounding words ”The”, ”cat”, ”on”, and ”the”.

In Skip-gram, the model would be trained to predict the surrounding words ”The”, ”cat”, ”the”, and ”mat” based

on the target word ”sat”. Both methods help in understanding the context of words and can be useful in natural

language processing tasks like text classiﬁcation, machine translation, and sentiment analysis.

In conclusion, Bag-of-Words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), and Word2Vec

are all widely used techniques in natural language processing for text representation. BoW represents a text

as a collection of words and their frequencies, but does not consider the context of the words. TF-IDF takes

(image courtesy: https://arxiv.org/pdf/1301.3781.pdf

Figure 4: CBOW & Skip-gram architectures.

3.3 Word2vec

into account the frequency of a word in a document and across all documents, but still does not capture the

semantic meaning of words. Word2Vec, on the other hand, represents words in a high-dimensional space

based on their context and semantic meaning, allowing for more accurate representations of words and their

relationships. While BoW and TF-IDF can be useful in certain applications, Word2Vec has shown superior

performance in various natural language processing tasks, including text classiﬁcation, sentiment analysis, and

machine translation. However, it requires more data and computing resources to train the model. Ultimately,

the choice of which technique to use depends on the speciﬁc task and available resources.

References

Feature Extraction and Embeddings in NLP: A Beginners guide to understand Natural Language Process-

ing,Analytics Vidhya,Siddharth M,July 2021

How to turn Text into Features, Tiago Duque,Towards Data Science,November 2020

Natural Language Processing — Feature Extraction Techniques, Rishi Kumar,Medium, August 2021

MOST POPULAR WORD EMBEDDING TECHNIQUES IN NLP Dataaspirant, August 2020

TF(Term Frequency)-IDF(Inverse Document Frequency) from scratch in python,Yassine Hamdaoui,

medium, December 2019

TF-IDF from scratch in python on a real-world dataset,William Scott, towardsdatascience, February 2019

About the Author

Jinsu Ann Mathew is a research scholar in Natural Language Processing and Chemical Informatics.

Her interests include applying basic scientiﬁc research on computational linguistics, practical applications of

human language technology, and interdisciplinary work in computational physics.

Part II

Astronomy and Astrophysics

Why the Sky is not the Limit for Astronomy

Research?

by Linn Abraham

airis4D, Vol.1, No.3, 2023

www.airis4d.com

The sky has attracted the curiosity of the human mind ever since the dawn of civilization and continues to

excite us to this day. Our concept of the sky changed drastically during the scientiﬁc revolution. From being

considered the abode of the Gods, it has become the cosmic laboratory to conduct observations and learn about

the universe. The scientiﬁc revolution was spearheaded by astronomers like Copernicus, Galileo, and Newton

amongst countless others, who were intrigued by the mysteries of the sky and dared to ﬁnd answers. In this

article, we shall try to answer some of the following questions. What is astronomy? What is the state of research

in astronomy? How has astronomical research beneﬁted the broader scientiﬁc community? Does research in

astronomy have any impact on the life of the common man apart from satisfying his curiosity? Is it still relevant

to pursue astronomy research today?

1.1 What is Astronomy?

The dictionary deﬁnes it as the study of celestial objects, space, and the physical universe as a whole. You

might be tempted to ask what these celestial objects are and why they are so important to study. The objects

most familiar to ancient humans were probably the Sun, the moon, and the stars. While stars moved in circular

patterns during the night, some of them showed eccentric motions. These wanderers were the planets and they

didn’t twinkle like the other stars. They were noticed by the ancient civilization like the Babylonians, Greeks

and others. In the early 20th century, when Hubble was able to measure the distance to a particular star known

as a Cepheid variable (M31 V1) he found it to be 2.5 million light years away from us. The average distance to

most others stars in the night sky is just around 5 light years. Thus we came to know of galaxies of stars beyond

our own. These galaxies look like fuzzy objects in the sky compared to the pinpricks of light that comes from

stars.

With the discovery of ground-based telescopes and telescopes sent to space on satellites, today the sky

has turned into a cosmic zoo. Several types of objects have since been discovered. There are Nebulae that are

the cradles of stars. Neutron stars and black holes, which are the favorites of science ﬁction writers. Brown

dwarfs and white dwarfs, Quasars and pulsars. The red giants and Cepheid variables and much more. Then

there are several astrophysical phenomena like solar and lunar eclipses, red moons and aurora, supernovae and

GRBs, solar ﬂares, comets, meteors, and meteorites. Several ﬁelds are currently under study in astrophysics.

These include space weather prediction, search for near-Earth objects, planetary explorations, stellar evolution,

1.2 Astronomy in the Ancient Era

Figure 1: The milky way galaxy as seen over Lake Tahoe (Image Credit: Tyler Clemmensen on Unsplash.com)

gravitational waves, dark matter, and dark energy amongst others.

But is it worthwhile to devote time and resources to studying such things when we have more important

problems to solve here on Earth? Why do people spend huge amounts of money and time on such frivolous

pursuits? In order to understand this, let us look at the history of astronomy and see what lessons it has to teach

us.

1.2 Astronomy in the Ancient Era

Astronomy is one of the oldest branches of science and one that has revolutionized Science in many ways.

Man’s curiosity about the sky probably started in pre-historic times. His pattern-seeking brain sought answers

to things like the motion of the stars, the waxing and waning of the moon, and the freckles on the lunar surface.

Answers were scarce to come by and he had to make do with stories and fables. But those who persisted in

their investigations did not do so in vain. Observations of the sky led them to discover a star that never changed

its position in the sky. This was the pole star or polaris that always pointed north. In hindsight now we know

the motion of the stars to be caused due to the rotation of the Earth about its axis. Hence a star that is directly

above the north pole would scarcely move. This helped them to navigate the vast oceans especially during the

night when the Sun is no longer there to provide directions. Observations of the lunar cycles, the solstices,

and equinoxes helped to keep time by creating calendars and helped them decide when to plant their crops.

Because of the tilt of the Earth’s axis the Northern hemisphere appears to point maximally towards the Sun once

during the year and away from the Sun at the end of the year. The subsolar point is the place which receives the

maximum intensity of the Sun’s light. This point moves North to South throughout the year through an angle

of 47°. Solstices designate the point in the Earth’s orbit around the Sun when the Sun’s path in the sky or the

subsolar point is the farthest North or South from the equator. The duration of day and night varies both with

the latitude of a place and the point where you have reached in the Earth’s orbit around the Sun. But at any point

on the Earth’s surface the Solstices are marked by the days that have longest and shortest durations.

1.3 The Age of Scientiﬁc Revolution

Figure 2: These towering tendrils of cosmic dust and gas sit at the heart of M16, or the Eagle Nebula. The

aptly named Pillars of Creation, featured in this stunning Hubble image, are part of an active star-forming region

within the nebula and hide newborn stars in their wispy columns. (Image Credit: NASA, ESA and the Hubble

Heritage Team (STScI/AURA))

Figure 3: Star trails captured using long exposure photography. (Image Credit: Michael Hull on Unsplash.com)

1.3 The Age of Scientiﬁc Revolution

Figure 4: Formation of the Seasons. (Image Credit: National Geographic)

Figure 5: Explanation of Solstices. (Image Credit: CosmoVerse Youtube Channel)

1.3 The Age of Scientiﬁc Revolution

Figure 6: Joseph von Fraunhofer (1778-1826) demonstrating the spectroscope

1.3 The Age of Scientiﬁc Revolution

The giants that spearheaded the scientiﬁc revolution like Aristotle, Copernicus, Galileo, Kepler, and

Newton were all astronomers. Science as we know it today emerged from the eﬀorts of these pioneers. The

Copernican hypothesis of a heliocentric universe was nothing short of an Earth-shattering event when it came to

be conﬁrmed through the observations of Galileo and others. Issac Newton framed his laws of motion drawing

heavily upon Kepler’s work on the laws of planetary motion.

Astronomy contributed heavily to the development of science in the centuries that followed. Fraunhoﬀer

who tried to analyze the Solar spectrum found dark lines in it. Scientists later found that these lines could

be used to identify the chemical composition of stars. The revolutionary technique of spectroscopy was born.

Using this we came to know that the bright blobs of light in the heavens, the Sun, and the stars are not solid or

liquid but made of gas, particularly hydrogen. Today spectroscopy is not only used in astronomy but in other

ﬁelds such as chemistry, material science, biology, and environmental science.

After nuclear physicists unlocked the power of the atom by nuclear ﬁssion processes it didn’t take them

long to identify that the huge source of power in the stars might be from a similar process. We know stars today

to be exploding hydrogen bombs that work by the principle of nuclear fusion.

Astronomy was used to measure the speed of light and also to test the theories of Albert Einstein. Most

famously his general relativity theory was tested by Arthur Eddington during the solar eclipse of 1919. Einstein’s

theory predicts that if light from a star passes very close to a massive object like the Sun, it can bend due to

the distortion of space-time. However, when the stars are in the line of sight with the Sun it is not possible to

photograph them due to the blinding light of the Sun. The test needed to wait till the next solar eclipse when

these stars could be precisely imaged and then compared to images of the same stars viewed from a diﬀerent

vantage point without the Sun in between. The precession of mercury’s orbit was also something that could

not be explained by Newtonian mechanics but was a natural prediction in Einstein’s theory. Many of his other

theories are such that they can only be tested by astrophysical observations.

1.4 Astronomy in The Space Race Era

Figure 7: The highest resolution image of the 1919 eclipse. It is the result of applying modern image

processing techniques — including image restoration, noise reduction, and removal of artifacts — to the

original photographic plate copy. It unveils stunning details in the solar corona, a giant prominence emerging

from the upper right part of the Sun, and stars in the constellation of Taurus (The Bull) that were used to conﬁrm

general relativity’s predictions.

Figure 8: A diagrammatic sketch of the theory behind the experiment.

1.4 Astronomy in The Space Race Era

Figure 9: NASA astronaut Kevin Ford, watches a water bubble ﬂoat freely between him and the camera,

showing his image refracted, in the Unity node of the International Space Station.(Image Credit: NASA)

1.4 Astronomy in The Space Race Era

The ﬁrst man-made satellite sent to orbit around the Earth had no other purpose other than to demonstrate

Soviet superiority. However, this event marked the beginning of the space race in which the United States and

the Soviet Union constantly competed with each other. All of humanity was to beneﬁt from this race. Soon the

potential of satellites was explored for studying and forecasting weather, and extreme events such as hurricanes

and tornadoes. They were used for data transmission and defense purposes. The GPS system that is heavily

used today was developed as part of the military applications of satellites. The moon that our ancestors could

only look at and weave stories about were touched upon by a man in the year 1969.

Space provides scientists with unique possibilities for conducting research. This includes providing an

environment with long-term exposure to microgravity, radiation, and vacuum conditions. It also provides a

unique vantage point for observing the universe. In space, the blurring eﬀects of the atmosphere that makes

stars twinkle and spread out are absent. The sky can be observed in its full glory in frequencies that would

normally be blocked by our atmosphere such as UV, X-rays, and Gamma Rays.

NASA the premier organization that conducts space and astronomy research has a web page dedicated to

spin-oﬀ technologies. These are technologies that have found commercial use in our daily lives. The page

now lists almost 2000 such technologies and includes some of the most important technologies of our age.

A few of these include solar cell, water ﬁltration based on activated charcoal, freeze-dried food, infra-red ear

thermometers, portable kidney dialysis machines, satellite television, advanced prosthetics, scratch-resistant

lenses, and memory foams.

1.5 The Impact of Astronomy Research

Analyzing the historical evidence it becomes clear that the importance of astronomy research to the world

cannot be overstated. The beneﬁts of astronomy research as in research in other fundamental sciences come in

1.6 The story behind the GPS and the Wireless Internet

a multitude of ways most of which are unexpected. It satisﬁes the curiosity inherent in human beings and opens

up the arena for bigger and tougher questions. This pushes the limits of our scientiﬁc knowledge and technology

and drives the quest for knowledge in several diﬀerent and related ﬁelds.

Research in astronomy calls for extreme engineering feats and innovation. There is probably a reason

why the usage ”at an astronomical scale” has become commonplace. Astronomy shows us the coldest possible

temperatures in the depths of space to the hottest possible temperatures in the cores of stars and galaxies. The

fastest moving things, the most energetic phenomenon in the universe, and the most massive objects in our

universe. The countless number of stars in our universe are greater than all the grains of sand on the Earth.

The development of technology demanded by astrophysical research drives economic growth as well and leads

to an overall improvement in our quality of life. The increased economic growth can further fuel research in

astronomy.

1.6 The story behind the GPS and the Wireless Internet

We saw several examples of how the technology that is often developed for astronomy research has

contributed to improvements in several related ﬁelds. However, let us take just two diﬀerent technologies that

easily demonstrate its economic and scientiﬁc value. The GPS network today is valued at billions of dollars not

to say anything about the value of countless lives it has helped save. The cesium atomic clocks abroad every

one of these satellites was developed exclusively to test the prediction of Einstein’s theory of relativity which

predicted a very small discrepancy between the time shown by a clock on the ground and one that is thousands

of miles above it and moving at very high velocities. Without accounting for these relativistic eﬀects, the GPS

system would accumulate position errors of several kilometers per day, making it practically useless.

The market for WiFi devices is estimated at hundreds of billions of dollars, not counting the economic

transactions expedited by its presence. John O’Sullivan was working on a method to reconstruct the signal of

a primordial black hole whose information was smeared out due to the interstellar medium that aﬀected the

low-frequency and high-frequency components of the signal diﬀerently. He designed a physical chip for doing

Fast Fourier Transforms (FFT) that later proved to be helpful when he was asked to work on a new project that

pioneered the use of low-power radio signals for wireless internet communication. Here the issue was with the

signals being smeared out due to the reverberations caused by a room’s geometry.

1.7 Astronomy in the Age of Big Data

Astronomy has always been at the forefront of technological innovations. Johannes Kepler can very well

be considered the ﬁrst data scientist. After the death of his mentor Tyco Brahe, Kepler inherited a wealth of data

that consisted of countless tables of measurements. Tyco Brahe used several pieces of equipment, including

quadrants and sextants, to measure the angles between celestial objects and the horizon. He noted the positions,

the time of observations, the weather conditions, and the quality of observations. It took Kepler more than six

years to formulate his ﬁrst law of planetary motion from this data. He arrived at the simplest model that ﬁt the

observed data, which was an elliptical orbit.

Today astronomy ﬁnds itself among the scientiﬁc endeavors that produce and analyze big data. There

are robotic telescopes that scan the entire sky multiple times each night taking pictures with short and long

exposures. The opening up of new regions of the electromagnetic spectrum including radio, infrared, UV, X-ray

and Gamma-ray and the discovery of gravitational waves have opened up the ﬂoodgates of data. Add to all of

1.8 References

Figure 10: The National Radio Astronomy Observatory’s Green Bank Telescope, Green Bank, West Vir-

ginia.(Image Credit-Britannica Encyclopedia)

these the digitization of archival data from old photographic plates and we have a ﬁeld that has the potential for

huge discoveries to be made.

Astronomers have been making use of machine learning and artiﬁcial intelligence techniques for solving

several challenging problems in astronomy. Most notable among them are the star-galaxy separation problem,

the galaxy classiﬁcation problem, and the detection of signals from gravitational wave data. With several

discoveries waiting to be made from existing data and several upcoming surveys already on the drawing board,

it is indeed an exciting time for research in astronomy.

1.8 References

1. Astronomy in Everyday Life

2. Joseph von Fraunhofer and the Solar Spectrum

3. Ole Roemer and the Speed of Light

4. What has astronomy done for you lately?

5. NASA Spinoﬀs

6. The man who made Einstein world-famous

7. Highest resolution image of the 1919 solar eclipse

8. Do stars move?

9. Seasons

About the Author

Linn Abraham is a researcher in Physics, specializing in A.I. applications to astronomy. He is

currently involved in the development of CNN based Computer Vision tools for classiﬁcations of astronomical

sources from PanSTARRS optical images. He has used data from a several large astronomical surveys including

SDSS, CRTS, ZTF and PanSTARRS for his research.

Variable Stars

by Sindhu G

airis4D, Vol.1, No.3, 2023

www.airis4d.com

2.1 What are variable stars?

Stars, the amazing objects in the night sky ! All of us have been curious about stars. In our childhood

days, all of us have watched stars in the night sky and wondered what stars are, what these objects are made

up of, how far away these stars are from us, how they produce light and so many such questions arises in our

minds. Stars in the night sky is a wonderful feast of sight that makes us happy. The relationship between stars

and human beings started from ancient times. We already know what stars are? They are luminous spheroids

of plasma held together by its own gravity. They are the fundamental building blocks of galaxies. When we

look into the sky, many of the stars appears to be constant in brightness. But most of these stars show at least

some variation in luminosity in their lifetime. But these brightness variations are more slowly than that can be

noticeable during one human lifetime. But some stars change their brightness over time with short periods. If a

star’s brightness as seen from Earth changes with time, then it is called a variable star. Figure 1 shows a variable

star RS Puppis.

Figure 1: RS Puppis (Image Courtesy: NASA, ESA, and the Hubble Heritage Team (STScI/AURA)-Hubble/Europe

Collaboration)

With the help of Figure 2 we can understand about the brightness changes of a variable star. The star in

the round shows some brightness at a time , say t1. When we watch it on another time t2, it is less brighter

2.2 Nomenclature

than as it is seen at t1. These brightness changes in variable stars can range from a thousandth of a magnitude

to twenty magnitudes. These changes in brightness can be from periods of a fraction of a second to years. All

these changes depend on the type of the star. There are many reasons for the variability of these stars. These

brightness variations may be due to the internal changes of the star or may be due to some external properties.

Brightness changes of Mira variable is given in Figure 3 and in Figure 4.

Figure 2: Variable Stars (Image Courtesy: AAVSO)

Figure 3: Mira variable close to maximum of brightness(Image Courtesy : AlltheSky.com)

Figure 4: Mira close to minimum of brightness(Image Courtesy : AlltheSky.com)

2.2 Nomenclature

When we discover a new variable star, that discovery must be submitted for oﬃcial certiﬁcation. Some

criteria for the certiﬁcation are given here. Deﬁnite evidence of variability must be given. The brightness of

the star must be determined reliably. The accurate position of the star must be given. Moreover the type of

variability of the star must be known. An independent observer will conﬁrm the variability of the star, especially

when the amplitude is small. After the certiﬁcation, the star is given an oﬃcial name. A German astronomer

2.3 Classiﬁcation

named Friedrich Wilhelm August Argelander began a careful study of variable stars.The modern system for the

nomenclature of variable stars was developed by Argelander. Stars with Greek-letter names or small-Roman-

letter names keep that name. The other stars are designated with one or two capital letters, followed by the

genitive form of the constellation’s Latin name. The ﬁrst discovered variable stars were designated with the

letters R to Z. For the next discovery the letters RR through RZ, SS through SZ, up to ZZ are used. Then

the letters AA through AZ, BB through BZ,and up to QQ through QZ (with J omitted) are used for the next

discovery of variable stars. This system gives 334 designations per constellation. When these combinations

are exhausted, the subsequent variables are numbered in the order of discovery, starting with the preﬁxed V335

onwards, followed by the genitive form of the constellation name. There are many other designations which can

be assigned to a variable star.

2.3 Classiﬁcation

Variable stars are divided into two main groups, intrinsic variables and extrinsic variables.

2.3.1 Intrinsic variables

Intrinsic variable stars show brightness changes because of the changes in the physical properties of the

stars. Intrinsic variables are further classiﬁed into pulsating variables, eruptive variables and cataclysmic

variables.

2.3.1.1 Pulsating variables

Pulsating variables periodically expand and contract their surface layers. They change their eﬀective

temperature, size, and spectral properties in this process. Pulsations may be radial or non-radial.

eg : Cepheids, Long Period Variables, RR Lyrae etc.

2.3.1.2 Eruptive variables

They experience eruptions on their surfaces because of violent processes and ﬂares occurring in their

chromospheres and coronae. eg : Herbig Ae/Be stars, giants, supergiants etc.

2.3.1.3 Cataclysmic variables

They undergo a cataclysmic change in their properties which is caused by thermonuclear processes either

in their surface layers or deep within their interiors. eg : supernovae, dwarf novae, symbiotic stars etc.

2.3.2 Extrinsic variables

In extrinsic variable stars the brightness variations are due to some external properties. Extrinsic variables

are classiﬁed into eclipsing binaries and rotating variables.

2.3.2.1 Eclipsing binaries

These stars periodically eclipse each other, causing a decrease in the apparent brightness of the system as

seen by the observer. eg : Algol variables, Beta Lyrae variables, W Ursae Majoris variables etc.

2.3 Classiﬁcation

2.3.2.2 Rotating variables

In these stars variability is caused by phenomena related to their rotation. eg : Ellipsoidal variables, BY

Draconis stars, SX Arietis variables etc.

In the next article we will explain more about the classifcation of variable stars.

References

Understanding Variable Stars, John R Percy, Cambridge University Press.

Variable star, Wikipedia

GCVS variability types, Samus N.N., Kazarovets E.V., Durlevich O.V., Kireeva N.N., Pastukhova

E.N.,General Catalogue of Variable Stars: Version GCVS 5.1,Astronomy Reports, 2017, vol. 61, No. 1,

pp. 80-88 2017ARep...61...80S

Variables: What Are They and Why Observe Them?

Types of Variable Stars: A Guide for Beginners

Variable Stars

What is a variable star?

Light Curves of Variable Stars

About the Author

Sindhu G is a research scholar in Physics doing research in Astronomy & Astrophysics. Her research

mainly focuses on classiﬁcation of variable stars using diﬀerent machine learning algorithms. She is also doing

the period prediction of diﬀerent types of variable stars, especially eclipsing binaries and on the study of optical

counterparts of x-ray binaries.

Part III

Biosciences

CRISPER Cas9 - The Gene Therapy

by Geetha Paul

airis4D, Vol.1, No.3, 2023

www.airis4d.com

In a biologist’s toolbox, CRISPR is a unique blend of powerful, relatively simple gene-editing tool which

allows scientists to modify, correct, or delete regions of DNA with high precision. In the past years we were

unaware of the root cause for the disease. Today, most medical therapies aim to treat diseases by targeting

abnormal proteins associated with the disease.These abnormal proteins occur due to mutations in speciﬁc genes

in a patient’s DNA. The 2020 Nobel Prize in Chemistry was awarded to CRISPR-Cas pioneers Emmanuelle

Charpentier and Jennifer Doudna, who developed the CRISPR-Cas system to edit genomic DNA precisely. This

technology pioneered at a breathtaking pace and is now used by almost every molecular biology laboratory

around the world in a myriad of organisms. The novel system called gene editing, the CRISPR- Cas9 technology,

took the lead in identifying the correct speciﬁc regions of DNA to treat the diseases. The Gene editing property of

CRISPR ﬁnds various applications in neurodegenerative disorders, genetic diseases, diagnostics and therapeutics

in cancer, blindness etc.

1.1 What is the CRISPR-Cas9 System ?

CRISPR stands for Clustered Regularly Interspaced Short Palindromic Repeats . A group of scientists,

including Dr. Emmanuelle Charpentier, discovered how to use this system as a gene-editing tool. Until then,

people knew “CRISPR” only as an acronym for the clustered regularly interspaced short ‘Palindromic Repeats’

(short sequences in a double stranded DNA or RNA molecule are identical and can be read in both directions as

repeats.Eg. ¿MALAYALAM¡) of genetic information that some bacterial species use as an antiviral mechanism.

The new CRISPR technique relies on two primary components: a Cas9, the enzyme that cuts the DNA

and a Guide RNA that guides the Cas9 precisely where in the genome to cut.

Cas9 represents the CRISPR-associated (Cas) enzyme that acts as “molecular scissors” to cut the DNA at

a location speciﬁed by a guide RNA.

Guide RNA (gRNA) is a type of RNA molecule that binds to Cas9 to form a complex and it speciﬁes the

location at which Cas9 will cut the DNA.

The whole process can be summarised as follows.

1. The ﬁrst step in CRISPR-Cas 9 technology is identifying the speciﬁc DNA sequence that needs modiﬁ-

cation. This is done by analysing the DNA of the studied organism and determining which gene or genes

are responsible for a particular trait or characteristic.

2. A guide RNA (gRNA) is designed so that it complements the target DNA sequence and binds to the target

DNA.

1.2 The Editing Process

The CRISPR- Cas9 gene editing process begins when the complex (Cas9 & guide RNA) recognizes the

short segment of DNA, adjacent to the target site.

3. The CRISPR-Cas system is delivered to the cells using a viral vector or through direct injection of the

CRISPR-Cas components.

4. Once the CRISPR-Cas system is delivered to the cells, the gRNA binds to the target DNA sequence,

guided by its complementary base pairing.

5. When the Cas enzyme is activated, it cuts the target DNA at a speciﬁc location. This initiates the

unwinding of the DNA helix, a double-strand break in the DNA and allows the guide RNA to pair with

the speciﬁc target sequence of DNA.

6. Every cell has an inbuilt natural DNA repair mechanism. It repairs the broken units in one of the two

ways: non-homologous end joining (NHEJ) or homology-directed repair (HDR).

7. The results of the CRISPR-Cas experiment are then analysed to determine whether the desired changes

were made to the DNA.

Figure 1: Illustration of the CRISPR- Cas9-mediated genome editing. Shows the single- guide RNA (sgRNA)

contains a protospacer recognising the target sequence followed by Protospacer Adjacent Motif (PAM). Forma-

tion of Non Homologous End Joining (NHEJ) and Homology Directed Repair (HDR).

Image source:https://www.sciencedirect.com/science/article/pii/S1931524415003321

In some cases Non Homologous End Joining (NHEJ) phases may occur, resulting in the addition or deletion

of few base pairs which disrupts the original DNA sequence and encodes gene inactivation.

A larger sequence of DNA can be separated using two diﬀerent guide RNA’s to target separate sites,on

either side of the desired deletion. Here the cleavage occurs to both sites and the deleted ends are repaired,

thereby deleting the intervening sequences.

Corrections to DNA can also be made by adding a DNA template to a Cas9 guide RNA complex. The

template is designed with sequences that exactly match with the DNA’s adjacent target site. The process called

Homology Directed Repair, the cell uses the template to repair the break. Thereby replacing the faulty DNA

sequences or even inserting a new gene.

1.3 Applications

Figure 2: Image source:https://crisprtx.com/gene-editing/crispr-cas9

DISRUPT- The single cut disrupts the original DNA. Formation of non-homologous ends.

DELETE - Deletion by two guideRNAs, After cleavage, non-homologous ends unite.

INSERT or CORRECT- Correction by CRISPR- Cas9 or Inserting a new gene, by a process called homology

directed repair.

Figure 3: Shows the Disruption and Deletion in DNA strand

Image source: https://crisprtx.com/gene-editing/crispr-cas9

Figure 4: Shows the Correction in DNA strand.

Image source: https://crisprtx.com/gene-editing/crispr-cas9

Overall, the steps involved in CRISPR-Cas9 technology are complex and require a high degree of precision,

but the potential applications of this technology are vast and it could revolutionise ﬁelds such as medicine,

agriculture, and environmental science.

1.3 Applications

In recent years, the CRISPR- Cas9 system has been increasingly used in cancer research and treatment and

remarkable results have been achieved.

Cancer is a refractory disease with high mortality and global attention. The malignant tumour causes 1 out

of 6 deaths globally thus threatening the lives of thousands of human beings. In basic research, CRISPR Cas9

can be used to study the function of speciﬁc genes that are involved in cancer development and progression. By

selectively editing or disabling genes, researchers can observe how these changes aﬀect cellular behaviour and

function, providing valuable insights into the underlying mechanisms of cancer.

In terms of diagnosis, CRISPR Cas9 can be used to create highly speciﬁc and sensitive diagnostic tests

for cancer. By introducing changes to speciﬁc genes or gene regions, the technique can be used to detect the

presence of cancer cells with high accuracy and sensitivity, making it a valuable tool in cancer screening and

diagnosis.

Finally, in therapy, CRISPR Cas9 can be used to develop new treatments for cancer. By selectively editing

genes that are involved in cancer development, researchers can potentially stop the growth and spread of cancer

cells, providing a highly targeted approach to cancer treatment. Additionally, CRISPR Cas9 can be used to

create new immunotherapies that enhance the immune response to cancer cells, providing a potential new avenue

for cancer treatment.

Overall, the CRISPR Cas9 gene editing technique has the potential to make a signiﬁcant impact in the

ﬁght against cancer and several other genetic disorders or mutations leading to various diseases, both in terms

of understanding the underlying mechanisms of the disease and developing new, targeted treatments. However,

more research is needed to fully explore the potential applications of this powerful technology in diagnosis and

therapy.

Collectively, these improvements proved the transformation of the CRISPR Cas9 editing tool from a blunt

instrument to a precision instrument in the Biologist tool box.

References:

CRISPR: development of a technology and its applications

Advances in therapeutic CRISPR- Cas9 genome editing

Discovery in CRISPR-Cas9 system

Application of the CRISPR- Cas9-based gene editing technique in basic research, diagnosis, and therapy

of cancer

CRISPR- Cas9 – a speciﬁc, eﬃcient and versatile gene-editing technology we can harness to modify,

delete or correct precise regions of our DNA

About the Author

Geetha Paul is one of the directors of airis4D. She leads the Biosciences Division. Her research

interests extends from Cell & Molecular Biology to Environmental Sciences, Odonatology, and Aquatic Biology.

Part IV

General

LiDAR Imaging

by Robin Jacob Roy

airis4D, Vol.1, No.3, 2023

www.airis4d.com

LiDAR stands for Light Detection and Ranging. It is fundamentally a ranging technology that uses light in

the form of a pulsed laser to measure distances and create 3D maps of the environment. LiDAR measures the

distance between source and target by hitting the target object with a laser beam and captures the return time

for the reﬂected light to reach the receiver.

A LiDAR system typically consists of a laser, a scanner, and a specialized GPS receiver. The laser emits

a pulsed beam of light, which is directed towards the target area. The scanner then sweeps the beam across the

target area, and the receiver detects the reﬂected light. By measuring the time it takes for the laser pulse to travel

to the target and back, LiDAR can determine the distance between the LiDAR system and the target.

LiDAR uses ultraviolet, visible, or near-infrared light to capture images of various objects. It has the

ability to detect a broad spectrum of materials such as non-metallic objects, rocks, rain, chemical compounds,

aerosols, clouds, and even individual molecules. For ground based surveys, a near-infrared laser is commonly

used, whereas a green wavelength laser is preferred for aquatic surveying operations. Green lasers are suitable

for coastal studies as they are visible to the naked eye and can penetrate shallow water. On the other hand,

near-infrared lasers are safe for the eyes, but they cannot reach beneath the water’s surface. Moreover, unlike

green lasers, water absorbs the energy from the near-infrared laser rather than reﬂecting it.

Figure 1: LiDAR System conﬁguration. Source: Travis S. Taylor, Introduction to Laser Science and Engineer-

ing.

1.1 LiDAR system components

Figure 2: Distance calculation using LiDAR. Source: Wikimedia Commons.

1.1 LiDAR system components

An airborne LiDAR comprises four key components that collaborate to generate precise and practical

outcomes:

LiDAR Sensors: During the ﬂight, LiDAR sensors scan the ground horizontally using pulses in green or

near-infrared bands.

GPS Receivers are utilized to track the airplane’s altitude and location, which are crucial for obtaining

precise terrain and elevation values.

Inertial Measurement Units (IMU): The airplane’s tilt is monitored by the Inertial Measurement Units

(IMU) during its travel, and this tilt information is used by the LiDAR systems to measure the incident

angle of the light pulse.

Data Recorders The pulse returns are recorded by a computer during LiDAR surface scanning, and this

data is converted into elevation.

Figure 1 shows a basic LiDAR system conﬁguration. The laser light pulses are split into two, just outside the

laser output coupler. One beam is directed to the receiver(typically a telescope with photon counting detectors)

and the other beam is directed towards a distant object. Using the split beam as the time reference, the reﬂected

beam’s return time is then compared and the diﬀerence between them will be twice the travel time required for

the reﬂected light to reach the receiver. Hence using the equation:

d =

c × t

the distance of the target object is calculated. Here d is the distance between the LiDAR and the target, c is the

speed of light and t is the time taken by the laser light to travel to the object or target, then travel back to the

detector. Figure 2 explains how the basic time-of-ﬂight principle is applied to laser range-ﬁnding.

1.2 Measuring trees with light

Light energy consists of photons. When these photons travel towards the ground, they encounter various

objects such as tree branches and bushes under the trees. Upon striking an object, some of the light is reﬂected

back towards the sensor. In case the object is small and has openings around it that enable the light to pass

1.3 Full Waveform vs Discrete LiDAR

Figure 3: An example LiDAR waveform returned from two trees and the ground. Source: NEON.

through, some of the photons move further down towards the ground. As a result of this, one pulse of light

may record multiple reﬂections due to some photons reﬂecting oﬀ objects like branches while others continuing

towards the ground.

The energy distribution returned to the sensor forms a waveform. The quantity of energy that is detected

by the LiDAR sensor is referred to as “intensity”. The peaks in the energy distribution correspond to the areas

where a greater number of photons or higher amount of light energy is returned to the sensor. These peaks

in the waveform often correspond to various objects on the ground, such as a branch, a cluster of leaves, or a

building. Figure 3 demonstrates LiDAR waveform returned from two trees and the ground.

1.3 Full Waveform vs Discrete LiDAR

A waveform or distribution of light energy is what returns to the LiDAR sensor. However, this return may

be recorded in two diﬀerent ways.

A Discrete Return LiDAR System records individual (discrete) points for the peaks in the waveform

curve. Discrete return LiDAR systems identify peaks and record a point at each peak location in the

waveform curve. These discrete or individual points are called returns. A discrete system may record 1-4

(and sometimes more) returns from each laser pulse.

A Full Waveform LiDAR System records a distribution of returned light energy. Full waveform LiDAR

data are thus more complex to process however they can often capture more information compared to

discrete return LiDAR systems.

1.4 Applications of LiDAR Imaging

Some of the major applications of LiDAR imaging are as follows:

FORESTRY: By utilizing liDAR-generated models of bare earth and tree canopy, it becomes feasible to

conduct detailed analyses of vast forested areas. Such models, along with others, can aid in approximating

parameters such as volume, density, mass, leaf area, canopy heights, and growth rate. The resultant

data can subsequently be employed to strategize logging, gauge productivity, chart the distribution of

biodiversity, pinpoint habitats of wildlife, and design programs for conservation. Individual tree crowns,

tree height, and density can be analyzed with LiDAR(Figure 4).

1.5 Advantages and Disadvantages

Figure 4: LiDAR can be used to analyse individual tree crowns, tree height and density. Source: Washington

State Department of Natural Resources.

NAVIGATION: LiDAR is now widely being used in autonomous navigation. The system use laser to

measure the speed and distance of all surrounding objects, allowing the vehicle’s computer to map its

surroundings in real time.

ARCHAEOLOGY: In the realm of archaeology, LiDAR technology can be employed to discover objects

concealed under the cover of forest canopies. LiDAR surveys can eﬀectively reveal prominent structures

that would be diﬃcult to detect from ground level, enabling archaeologists to identify sites that may have

otherwise gone unnoticed.

AGRICULTURE: In agriculture, LiDAR technology can be used to determine slope angle and direction

of the agricultural area. These factors help in the calculation of possible crop yield, pinpointing optimal

quantity and spots for fertilizer placement to maximize crop yield, and in planning eﬃcient drainage.

HYDROLOGY: Hydrologists delineate stream orders and tributaries from LiDAR.

1.5 Advantages and Disadvantages

LiDAR technology has several advantages and disadvantages, which are discussed below:

Advantages

1. High accuracy and precision: LiDAR systems can provide highly accurate and precise data, especially

when it comes to measuring elevation or distance. They can provide measurements with an accuracy of

a few centimeters or less.

2. Wide coverage area: LiDAR technology can cover large areas quickly and accurately, making it ideal for

mapping applications. It can provide detailed data over large areas that would be time-consuming and

costly to collect using traditional surveying methods.

3. Flexibility: LiDAR can be used in a variety of applications, including forestry, agriculture, urban planning,

and autonomous vehicles. It can provide 3D point cloud data that can be used for many diﬀerent purposes.

4. Safe: LiDAR systems are relatively safe as they use non-ionizing radiation, which does not harm human

beings or the environment.

1.5 Advantages and Disadvantages

Disadvantages

1. Cost: LiDAR systems can be expensive, especially when compared to other surveying methods. The cost

of data processing and analysis can also add to the overall cost.

2. Limited penetration: LiDAR systems are not suitable for penetrating thick vegetation or certain types of

surfaces such as water or ice, which can limit their use in some applications.

3. Weather dependency: LiDAR systems can be aﬀected by weather conditions such as rain, fog, or snow,

which can reduce their accuracy and eﬀectiveness.

4. Data processing: LiDAR data processing and analysis require specialized software and skilled profes-

sionals, which can add to the overall cost and time needed to generate useful information.

References:

Introduction to LiDAR

What is LiDAR

Light Detection and Rangind

LiDAR Facts

Travis S. Taylor (2020). Introduction to Laser Science and Engineering, 240. CRC Press.

About the Author

Robin is a researcher in Physics specializing in the applications of machine learning for remote

sensing. He is particularly interested in using computer vision to address challenges in the ﬁelds of biodiversity,

protein studies, and astronomy. He is currently working on classifying satellite images with Landsat and Sentinel

data.

Part V

Fiction

Curves and Curved Trajectories

by Ninan Sajeeth Philip

airis4D, Vol.1, No.3, 2023

www.airis4d.com

This article is a thought experiment on a virtual world with imaginary creatures capable of doing ﬁctitious

activities.

Episode 4

Characters:

1. Dot: a zero-dimensional creature

2. Angel the Light: a one-dimensional Angel

3. Mittu: a two-dimensional ant

4. Albert: a 3D boy.

Mittu was still ﬁnding it hard to follow the spacial dimensions that Albert was trying to explain. Albert

was immersed in his thoughts on explaining all of this to Mittu.

”Are you aware that you are a 3D species?” Albert broke the silence with his question to Mittu.

”What do you mean? I do not even know what this 3D is all about!” replied Mittu.

”Oh yes, you are. But because your brain never had the chance to experience a 3D realisation, it is reduced

to the 2D world. It is only because of your mental inability to visualise 3D that you ﬁnd it so intimidating”,

explained Albert.

”That’s even wired!”, exclaimed Mittu. ”How can my mental inability ever obscure reality?” He asked.

”Oh yes, it can”, said Albert. ”You might have heard about this whole story of dark matter that worries

scientists. We know that it is there. But unfortunately, there is nothing we know of that they have that our sense

organs can detect! Our sensorium’s capability limits us. That is a fact.”

“Hmm”, murmured Mittu. ”So what are you getting at?” he asked.

”Well, I am curious to know how 4D will appear to people like me who can’t visualise beyond 3D”, said

Albert.

”Oh, I see.. So how are you planning to do it?”, asked Mittu.

”I do not know! But I am sure that our ability to extrapolate logically in the geometric, rather mathematical

formalism of logic will help us to break into the unknown territory,” said Albert.

”You mean the Minkwisky and Remanian space-time that General Relativity talks about?” clariﬁed Mittu.

”I don’t know! I am just wishful,” admitted Albert. ”But since you have brought up General Relativity,

something has always puzzled me there. If you consider the Einstein Field equation

µυ

+ g

µυ

Λ =

8πG

µυ

and forget about the cosmological constants and assume that the constants are unity, it turns out to be G

µυ

= T

µυ

Here, the left-hand side is curvature of space-time, and the right-hand side is matter or energy.” said Albert.

”Yes, we all know that”, said Mittu.

”Yes, we all know it,” said Albert. ”But I do not understand the need for the matter to create curvature in

space-time. Can’t the reverse be true where space-time curvature manifests as Matter or Energy?” Albert put

forward a tricky question for Mittu to grasp.

”I am not getting you”, complained Mittu.

”All right, I may be wrong, but have you ever noticed distorted shadows on the ground when you walk

around?” Asked Albert.

”Oh yes!” exclaimed Mittu. ”When the sun rose, I noticed my shadow on the ground growing smaller as

the sun goes up. Depending on where I stand, sometimes part of my shadow vanishes or gets modiﬁed until it

is out of certain regions. I was fascinated by those and think of it as some magic property of certain ground

regions!” explained Mittu.

”Wonderful! That’s great observation”, said Albert. ”You are right. It is the magic of the ground. That

magic is the curvature of the surface. Since Light travels in straight lines, wherever there is a curvature, the

images get distorted.”, clariﬁed Albert. However, Mittu needed more convincing argument. So he asked: ”I am

not very convinced by what you say. My ground is ﬂat, and there is no curvature!”

Albert raised his eyebrows and continued, ”It is not that there is no curvature. Since your brain sees

everything as ﬂat 2D space, you cannot visualise it. You see only the projection from the 3D curvature on your

surface!”

”Oh, My! You are challenging the ontological existence of modiﬁed shadows! I am confused,” said Mittu.

”I am more concerned about the ontological existence of matter!” said Albert. ”If space-time curvature

can create the notion of matter, the world is much better!”

”Wait, wait. . . ” interrupted Mittu. ”Matter is Energy. How can curvature create it”?

”Good point”, nodded Albert. ”But how did you conclude that space-time has no inherent Energy? The

Field equation demands it”.

”But where did it come from?” questioned Mittu

”From the same source all the matter came from!” smiled Albert. ”See, I do not have an answer to your

question. It is beyond my comprehension. But if space-time curvature can bring in the notion of matter, you

don’t need a big bang or inﬂation to create the universe. Simple quantum ﬂuctuations can create space-time

vibrations and result in microscopic particle creation! ”

Mittu stared at Albert and said, ”That’s a long shot! I don’t think that I can agree.”

”Yes, yes, it is”, agreed Albert. ”Let us experiment”, continued Albert while picking up Mittu and placing

him on one of the arms of a turbine in his set-up. ”Walk around’ he said. Mittu went left and right, up and

down, when Albert asked, ”Do you ﬁnd anything peculiar?” ”No,” said Mittu. Albert started turning the turbine

at a constant speed. Mittu cried out, ”Oh, something is going on. I am feeling pulled down!” He said. ”Yes, you

are!” exclaimed Albert ”because you are moving in that direction!” he continued. He then stopped the turbine

and landed Mittu on the ground. ”How do you do that?” asked Mittu in excitement. Albert took Mittu on the

platform of a toy lift and switched it on to move upwards. ”Oh,” Mittu said when the ride started. When he

became silent, Albert asked, ”How do you feel now?”

”Nothing special! There was something similar to the earlier experience, but now everything is normal!”

he conﬁrmed. Albert switched oﬀ the lift. ”Ho!” said Mittu. ”I felt something similar now, the same way I had

initially, but with a diﬀerence. It looked exactly opposite to my initial feeling,” said Mittu.

”Excellent observations Mittu and thank you for your corporation” Albert took Mittu out from the lift and

said. ”In one, you experienced a constant pull in one direction all the time while in the other, it was there

initially and ﬁnally but were opposite in nature, right?” Albert tried to clarify his experience with Mittu.

”Yes, yes”, conﬁrmed Mittu.

”Do you remember the experience our friend Light had when you explained to him that it is because he is

moving along a curved path?” asked Albert.

”But he was indeed moving along a curved path, but here I was, just idling on the ﬂoor and still experiencing

all of it” Mittu raised his objection.

”No, no. You were not! In the ﬁrst case, you were on a curved path and in the second on a linear one!”,

said, Albert. ”But to you, both appeared identical and indistinguishable, simply due to your inability to visualise

the third dimension!” he clariﬁed.

”Oh, that’s strange,” said Mittu.

”Yes indeed, it is strange!” said Albert.

”But why did you do it? What has it got to do with the curvature and matter we have been discussing?”

asked Mittu.

”Oh yes, I am coming to that,” said Albert. ”I told you about the four-dimensional space. Since you nor I

can visualise it, let us call it a 4D manifold. I see only the 3D projection that remains inert, like the ﬂoor you

saw. Agreed?” he asked Mittu.

”Agreed!” aﬃrmed Mittu.

”Now assume that this 4D manifold is spinning toward the 4th axis. Then I should experience the same

kind of feeling that you had, like being pulled irrespective of the three directions that I can visualise or move

about,” said Albert while Mittu’s eyes twinkled like the night star.

”Pulling means a force, and my basic understanding of classical mechanics immediately demands the

existence of two entities for force to exist, which are mass and acceleration” Albert paused to see if Mittu

had anything to say. But Mittu was mesmerised. He has been driven into that unimaginable manifold of

mathematical logic.

”But my manifold and the 3D space are empty without any matter. So this is an acceleration that exists in

empty space!” said Albert

”So are you suggesting a curvature without any matter in your 4D constructed space-time?” asked Mittu

curiously.

”Absolutely!” said Albert. ”Let us call this curvature of our empty space lambda as we do not have a matter

associated with it yet!” he said.

”Wow! That’s interesting,” said Mittu. ”You are telling me of space-time with an inherent curvature and

accelerating for no known reason! Then what?” he asked.

”Yes, yes.. but it is not an empty space-time for me. To make my physics right, I call the equivalent matter

for this accelaration, the total matter of the universe! Since it is a constant, the conservation laws come into

play. Since I experience it and continue to experience it, it also brings in the notion of time” said Albert.

”Wow!” exclaimed Mittu

”To summarise, my 3D world has two entities; one is acceleration, and the other is time. This notion of

time is independent of the accelaration as both are orthogonal to each other (analogus to centripetal force and

tangential direction of velocity). This allows me to rewrite time in the dimension of distance by multiplying it

with some universal constant, say C” said Albert.

”Interesting!” said Mittu.

”However, the universe can’t remain like that. It is subject to quantum ﬂuctuations that build up inﬁnitesimal

curvatures in space-time, damaging its homogeneity!’ said Albert.

”Oh!” exclaimed Mittu. ”What’s that going to do?” He asked.

”I don’t know!” confessed Albert. ”But keeping the conservation laws intact, I can envisage three pos-

sibilities. There could be some ﬂuctuations that can locally cancel the original curvature of the space-time to

introduce inhomogeneity. Then there can be two others that either locally increases or decrease the space-time

curvature giving a positive or negative additional curvature to space-time!” he paused.

”Oh my, are you telling me that this neutralising curvature locally creates matter and the other two, in

addition, add a negative and positive attractive or repulsive force to the matter?” Mittu asked with both eyes and

mouth wide open.

”Honestly, I do not know,” confessed Albert for the second time. ”But this puzzle is killing me!” he said,

pulling backwards on his two arms ﬁxed to the ground.

”But, they say that the universe is expanding and that there is what they call a scale factor; what do you

say?” asked Mittu.

”Oh, if you can agree up to this point, that is simple”, said Albert. ”You are revolving in a dimension

that you cannot visualise. Even when you are all travelling the same angular distance θ, those far from you

cover a linear distance of rθ where r is the distance from you. But since you visualise this distance only as

the projection in your 3D world, it appears that the objects are moving away proportional to their distance from

you! Your position has no reference in the 3D world. So it appears immaterial in which direction you look; it

will all be moving away!” Albert said very casually.

There was absolute silence for a while. Nobody knows how long. Then Albert suddenly recollected that

he had to leave before someone came in search of him and thus stood up. He thanked Mittu, Light and Dot,

who had all fallen half asleep during the disclose as he walked away.

About the Author

Professor Ninan Sajeeth Philip is a Visiting Professor at the Inter-University Centre for Astronomy

and Astrophysics (IUCAA), Pune. He is also an Adjunct Professor of AI in Applied Medical Sciences [BCMCH,

Thiruvalla] and a Senior Advisor for the Pune Knowledge Cluster (PKC). He is the Dean and Director of airis4D

and has a teaching experience of 33+ years in Physics. His area of specialisation is AI and ML.

About airis4D

Artiﬁcial Intelligence Research and Intelligent Systems (airis4D) is an AI and Bio-sciences Research Centre.

The Centre aims to create new knowledge in the ﬁeld of Space Science, Astronomy, Robotics, Agri Science,

Industry, and Biodiversity to bring Progress and Plenitude to the People and the Planet.

Vision

Humanity is in the 4th Industrial Revolution era, which operates on a cyber-physical production system. Cutting-

edge research and development in science and technology to create new knowledge and skills become the key to

the new world economy. Most of the resources for this goal can be harnessed by integrating biological systems

with intelligent computing systems oﬀered by AI. The future survival of humans, animals, and the ecosystem

depends on how eﬃciently the realities and resources are responsibly used for abundance and wellness. Artiﬁcial

intelligence Research and Intelligent Systems pursue this vision and look for the best actions that ensure an

abundant environment and ecosystem for the planet and the people.

Mission Statement

The 4D in airis4D represents the mission to Dream, Design, Develop, and Deploy Knowledge with the ﬁre of

commitment and dedication towards humanity and the ecosystem.

Dream

To promote the unlimited human potential to dream the impossible.

Design

To nurture the human capacity to articulate a dream and logically realise it.

Develop

To assist the talents to materialise a design into a product, a service, a knowledge that beneﬁts the community

and the planet.

Deploy

To realise and educate humanity that a knowledge that is not deployed makes no diﬀerence by its absence.

Campus

Situated in a lush green village campus in Thelliyoor, Kerala, India, airis4D was established under the auspicious

of SEED Foundation (Susthiratha, Environment, Education Development Foundation) a not-for-proﬁt company

for promoting Education, Research. Engineering, Biology, Development, etc.

The whole campus is powered by Solar power and has a rain harvesting facility to provide suﬃcient water supply

for up to three months of drought. The computing facility in the campus is accessible from anywhere through a

dedicated optical ﬁbre internet connectivity 24×7.

There is a freshwater stream that originates from the nearby hills and ﬂows through the middle of the campus.

The campus is a noted habitat for the biodiversity of tropical Fauna and Flora. airis4D carry out periodic and

systematic water quality and species diversity surveys in the region to ensure its richness. It is our pride that

the site has consistently been environment-friendly and rich in biodiversity. airis4D is also growing fruit plants

that can feed birds and provide water bodies to survive the drought.