Cover page

Image Name: Tiny Galaxies - stellar clusters.

Open Cluster NGC 376,This image from the NASA/ESA Hubble Space Telescope captures a small portion of

the Small Magellanic Cloud (SMC). The SMC is a dwarf galaxy and one of the Milky Way’s nearest neighbors,

lying only about 200,000 light-years from Earth. It makes a pair with the Large Magellanic Cloud, and both

objects are best seen from the Southern Hemisphere, but are visible from some northern latitudes as well. Read

more here: https://science.nasa.gov/missions/hubble/hubble-captures-a-glittering-neighbor/

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.12, 2023

www.airis4d.com

Welcome to the 12th edition of airis4D, Vol.1, 2023! We have completed the ﬁrst volume of airis4D.

This edition delves into diverse topics like astronomical phenomena, biological age, vascular aging, and the

intricacies of workaholism and depression. Each article oﬀers insightful perspectives and knowledge on these

intriguing subjects, enriching our understanding of complex aspects of science, health, and human behavior.

”Introduction to Quantum Machine Learning - Part 3” by Blesson George presents a comprehensive dive

into quantum computing concepts, focusing on qubits, quantum circuits, and the intriguing model of adiabatic

quantum computing. The article delves into the adiabatic theorem’s role in time-dependent Hamiltonians, oﬀer-

ing an alternative quantum computation approach. It explores the meticulous process of encoding, preparing,

annealing, and measuring within the adiabatic model, showcasing its potential for eﬃcient problem-solving.

Furthermore, the article introduces quantum annealing as an optimization technique leveraging qubits’ unique

traits for complex problem landscapes. Both adiabatic quantum computing and quantum annealing emerge as

promising frontiers in the quantum computing realm, presenting innovative solutions across diverse domains as

quantum technologies advance.

”Multimodal NLP: Understanding the World Like Us” by Jinsu Ann Mathew introduces the concept of

Multimodal Natural Language Processing (NLP) in Artiﬁcial Intelligence (AI), aiming to mimic human senses

for a comprehensive understanding of information. The article explores how traditional AI focuses on single

modalities like pictures or text, whereas multimodal AI combines information from diﬀerent sources such as

images, text, sound, and smell to grasp a richer context. It elucidates the working of multimodal AI through

its Input, Fusion, and Output modules, drawing parallels to human sensory perceptions. The Fusion Module,

likened to the brain, integrates data from various modalities through techniques like early fusion, alignment-

based fusion, and late fusion. The Output Module translates this fused information into understandable or

actionable outcomes, mirroring human interaction. Multimodal NLP represents a signiﬁcant leap in creating

AI systems that engage with humans in a manner akin to our complex communication, signaling a promising

stride toward more intelligent and versatile machines.

”Integrated Gradients: A Model Interpretability Technique for Machine Learning” by Linn Abraham

introduces a method for developing Interpretable Machine Learning models. Usually, machine learning models

are considered black boxes. He discusses ways in which this limitation may be overcome.

In his ﬁfth article in the series, ”Black Hole Stories – 5, Some Concepts in Einstein’s General Theory of

Relativity”, Professor Ajit Kembhavi narrates the fundamentals for understanding General Relativity through

simple illustrations. The introduction of Special Relativity brought many changes in our understanding that

need to be clear when we study General Relativity. This article provides those details.

”Exploring the Enigmatic World of Be/X-ray Binaries” dives into high-mass X-ray binaries, focusing

on Be/X-ray binaries, where a massive Be star interacts with a neutron star or black hole. These binaries,

characterized by their intense X-ray emissions, exhibit intriguing dynamics between the Be star’s circumstellar

disk and the compact object. The article details the unique traits of Be stars, their circumstellar disks, and the

neutron stars’ dense nature, explaining the X-ray emission bursts resulting from the interaction between the

neutron star and the disk. It highlights the transient behavior of these binaries, providing valuable insights into

accretion physics, mass transfer dynamics, and neutron star properties. Additionally, it outlines observational

techniques—X-ray, optical, infrared, and radio astronomy—used to study these systems, emphasizing the

signiﬁcance of Be/X-ray binaries as cosmic laboratories for understanding extreme astrophysical conditions and

fundamental processes in the cosmos. The article serves as an introduction to forthcoming explorations into

various categories of Be/X-ray binaries, promising deeper insights into these intriguing astrophysical systems.

The article ”Exploring the Complexities of Ageing Process with respect to Biological age and Chronological

age” investigates ageing through the lenses of biological and chronological age. It emphasizes that while

chronological age measures time since birth, biological age considers genetic, lifestyle, and physiological factors,

providing a more accurate gauge of ageing. The exploration of ageing biomarkers like telomere shortening,

DNA methylation, and vascular ageing contributes to understanding the ageing process. The distinction between

biological and chronological age has implications for health outcomes, interventions, and legal considerations.

Epigenetic studies oﬀer promising avenues for accurately calculating biological age, impacting discussions about

legal age changes and aligning them with biological age. Biomarkers associated with vascular ageing, such

as arterial stiﬀness, endothelial function, and inﬂammatory factors, oﬀer insights into cardiovascular disease

risks in ageing individuals. Further research on these biomarkers is essential to improve clinical relevance and

enhance predictions related to vascular ageing-related conditions. Understanding these complexities provides

opportunities for interventions, longevity, and healthy ageing.

The article ”Workaholism and Depression” explores the diﬀerence between healthy hard work and detri-

mental workaholism, focusing on their impact on mental health and personal well-being. It distinguishes

between self-motivation that fosters healthy hard work and excessive workaholism that leads to stress, low self-

esteem, and depression. It highlights societal pressures that can push individuals into unsuitable professions,

resulting in stress and obsession, leading to depression. The piece emphasizes the withdrawal from social life

by workaholics and the adaptability of hard workers in various aspects of life. Additionally, it discusses the

genetic links between mood disorders and clinical depression and emphasizes the health impacts of workaholism

compared to the balanced approach of hard workers. The article encourages introspection to discern between

stressful work and healthy hard work, emphasizing passion and internal drive as crucial factors. In conclusion,

it cites Thomas Edison’s approach as an example, showcasing his willingness to learn from failures—a trait of

hard workers who view failures as opportunities for growth.

As we conclude the inaugural year of airis4D Journal, we extend our heartfelt gratitude to our readers,

contributors, and supporters. This year has been a journey of exploration, discovery, and learning. We’re

committed to continuing our pursuit of knowledge and sharing insightful content across diverse domains in the

years ahead. Thank you for being part of our successful journey, and here’s to many more years of insightful

discoveries and enlightening discussions with airis4D.

iii

Contents

Editorial ii

I Artiﬁcial Intelligence and Machine Learning 1

1 Introduction to Quantum Machine Learning - Part 3 2

1.1 Adiabatic Quantum Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Quantum Annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Multimodal NLP : Understanding the World Like Us 5

2.1 What is Multimodal NLP? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Working of Multimodal AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.3 Input Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.4 Fusion Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.5 Output Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Integrated Gradients: A Model Interpretability Technique for Machine Learning 10

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3.3 Need for a Baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.4 Why not just use Gradients? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.5 Inspiration from Game Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.6 Evaluation of Attribution Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.7 Integrated Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

II Astronomy and Astrophysics 15

1 Black Hole Stories – 5

Some Concepts in Einstein’s General Theory of Relativity 16

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.2 The Schwarzschild Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.3 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.4 Types of Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.5 Conserved Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2 Exploring the Enigmatic World of Be/X-ray Binaries 21

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.2 Understanding Be/X-ray Binaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.3 Observational Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.4 Importance of the Study of Be High-Mass X-ray Binaries . . . . . . . . . . . . . . . . . . . . 24

CONTENTS

III Biosciences 26

1 Exploring the Complexities of Ageing Process with respect to Biological Age and

Chronological Age 27

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

1.2 Telomere Shortening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

1.3 Epigenetic Regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

1.4 DNA Methylation and Histone Modiﬁcations . . . . . . . . . . . . . . . . . . . . . . . . . . 29

1.5 CpG Islands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

1.6 Biomarkers of Biological Vascular Aging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

IV General 34

1 Workaholism and Depression 35

1.1 The Myths about Workaholism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

1.2 Fundamental Diﬀerences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Part I

Artiﬁcial Intelligence and Machine Learning

Introduction to Quantum Machine Learning

- Part 3

by Blesson George

airis4D, Vol.1, No.12, 2023

www.airis4d.com

In our previous article, we embarked on a detailed exploration of the fundamental concepts of quantum

computing surrounding qubits and quantum circuits. Delving deeper into the realm of quantum computing, let’s

augment our understanding of these intriguing subjects with a more nuanced discussion. Qubits, the quantum

counterparts to classical bits, exhibit a remarkable property known as quantum superposition, allowing them to

exist in multiple states simultaneously. This unique characteristic forms the bedrock of quantum information

processing, extending the possibilities of computation beyond the conﬁnes of classical binary systems. Moreover,

our examination of quantum circuits revealed them as the building blocks of quantum computation, where gates

such as the NOT gate and Hadamard gate play pivotal roles in manipulating and entangling qubits. In this

article, we discuss about an alternative model called adiabatic quantum computing, which manipulates states

using Hamiltonian operator.

1.1 Adiabatic Quantum Computing

1.1.1 Adiabatic Theorem

According to the quantum adiabatic theorem, a quantum system that begins in the non-degenerate ground

state of a time-dependent Hamiltonian will remain in the instantaneous ground state provided the Hamiltonian

changes suﬃciently slowly.

The solution to the time-independent Schr

odinger equation provides a straightforward solution when

dealing with a time-independent Hamiltonian.

|ψ(t)⟩ = H|ψ(t)⟩

with the initial quantum state given by

|ψ(t)⟩ = exp(−iHt)|ψ(0)⟩

However, the complexity of the problem increases signiﬁcantly when the Hamiltonian is time-dependent. In

such cases, the evolution of the system becomes more intricate.

In adiabatic quantum computing model, a computation is deﬁned by two Hamiltonians, namely H

initial

and H

final

, where a Hamiltonian is essentially a Hermitian matrix. The ground state of H

initial

, characterized

by the eigenvector with the smallest eigenvalue (also referred to as the ground state), must be a readily preparable

1.2 Quantum Annealing

state, such as a tensor product state.

The result of the adiabatic computation is the ground state of the ﬁnal Hamiltonian, H

final

. Consequently,

we select an H

final

whose ground state represents the solution to the given problem. An essential requirement

is that the Hamiltonians are local, meaning they involve interactions only among a constant number of particles.

This constraint can be likened to allowing gates that operate on a constant number of qubits in the standard

model. This condition ensures a concise classical description of the Hamiltonians by listing the matrix entries

of each local term.

The runtime of the adiabatic computation is dictated by the minimal spectral gap of all the Hamiltonians

along the straight line connecting H

initial

and H

final

, denoted as H(s) = (1 − s)H

initial

+ sH

final

Within the adiabatic model, several key steps unfold, each playing a crucial role in the computational

process:

1. Encode Your Problem (Boolean SAT Problem): In the initial phase, the problem at hand is encoded,

speciﬁcally in terms of a Boolean satisﬁability (SAT) problem. This step involves formulating the logical

conditions of the problem in a way that aligns with the Boolean framework.

2. Prepare Initial State of Qubits (Program Your Problem): Following the encoding, the next step involves

the preparation of the initial state of qubits. This is akin to programming the quantum system with the

information encapsulated in the Boolean SAT representation, setting the stage for subsequent quantum

operations.

3. Annealing Process (Slowly Change from Initial to Final State): The heart of the adiabatic computation

lies in the annealing process. Here, a gradual transformation occurs, moving the quantum system from

its initial state to the ﬁnal state. This gradual evolution is essential for maintaining adiabaticity, ensuring

that the system remains in its ground state throughout the process.

4. Measure Your Answer: Finally, the computational outcome is determined through the measurement of

the quantum system. The measured result provides the solution to the encoded problem, and the quantum

nature of the system allows for the exploration of multiple possibilities simultaneously, enhancing the

eﬃciency of certain computations.

1.2 Quantum Annealing

Quantum annealing is a computational paradigm that leverages principles from quantum mechanics to

tackle complex problem-solving tasks. At its core, it is an optimization technique designed to ﬁnd the most

eﬃcient solution to a given problem. Unlike classical computing, which relies on bits, quantum annealing

employs qubits—quantum bits—that can exist in multiple states simultaneously through superposition.

The process begins by encoding the problem into a set of parameters that can be represented in the language

of quantum mechanics. These parameters are then mapped onto qubits, conﬁguring the quantum system to

embody the problem space. Quantum annealing distinguishes itself through its unique annealing process, a

gradual transformation from an easily preparable initial state to a ﬁnal state that encodes the solution sought.

During annealing, the quantum system explores various conﬁgurations, adapting to the problem’s land-

scape. This exploration beneﬁts from the phenomenon of quantum tunneling, where the system can overcome

energy barriers that might impede classical algorithms. The adiabatic theorem ensures that the system evolves

in a way that preserves its ground state, allowing it to settle into the optimal solution as the annealing progresses.

The ultimate outcome is obtained through measurement, revealing the state of the qubits and thus pro-

viding the solution to the encoded problem. Quantum annealing holds promise for addressing combinatorial

optimization problems, such as those encountered in logistics, ﬁnance, and artiﬁcial intelligence, where the sheer

1.3 Summary

complexity of potential solutions makes traditional computational methods less eﬃcient. As quantum comput-

ing technologies advance, quantum annealing stands as a captivating avenue for exploring novel approaches to

problem-solving in diverse ﬁelds.

1.3 Summary

Our exploration of quantum computing has delved into the foundational principles of qubits, quantum cir-

cuits, and the distinctive model of adiabatic quantum computing. The adiabatic theorem’s guidance through time-

dependent Hamiltonians underscores the model’s potential as an alternative quantum computation paradigm.

The meticulous orchestration of encoding, preparation, annealing, and measurement within the adiabatic model

reﬂects its promise for eﬃcient problem-solving. Additionally, our investigation into quantum annealing has

unveiled a powerful optimization technique, leveraging the unique attributes of qubits for complex problem

landscapes. Both adiabatic quantum computing and quantum annealing stand as promising frontiers in the

quantum computing landscape, oﬀering innovative solutions to intricate problems as quantum technologies

continue to advance. The ability of quantum systems to explore multiple states simultaneously and overcome

classical computational constraints opens doors to transformative approaches in problem-solving across diverse

domains.

References

[1] Wittek, P. (2014). Quantum machine learning: what quantum computing means to data mining. Academic

Press.

[2] Prof.Andrew Childs - Lecture on the quantum adiabatic theorem -

https://www.cs.umd.edu/amchilds/teaching/w08/l18.pdf

[3] D. Aharonov, W. van Dam, J. Kempe, Z. Landau, S. Lloyd and O. Regev, ”Adiabatic quantum computation

is equivalent to standard quantum computation,” 45th Annual IEEE Symposium on Foundations of Computer

Science, Rome, Italy, 2004, pp. 42-51, doi: 10.1109/FOCS.2004.8.

About the Author

Dr. Blesson George presently serves as an Assistant Professor of Physics at CMS College

Kottayam, Kerala. His research pursuits encompass the development of machine learning algorithms, along

with the utilization of machine learning techniques across diverse domains.

Multimodal NLP : Understanding the World

Like Us

by Jinsu Ann Mathew

airis4D, Vol.1, No.12, 2023

www.airis4d.com

Think about how we experience things - we see, hear, touch, smell, and taste. Just like that, Artiﬁcial

Intelligence (AI) is trying hard to understand the world just like we do with all our senses. This eﬀort is called

multimodal Natural Language Processing (NLP), and it’s a really promising idea. Just like we use all our senses

at once, AI wants to do the same. It wants to combine information from what it sees, hears, and reads to

understand conversations better. This will make AI smarter and able to understand us better.

Imagine you’re trying to teach an AI to understand what’s happening in a kitchen. Traditional AI might

focus only on pictures or only on written descriptions. Let’s say you show it a picture of a pot on the stove.

Traditional AI might recognize the pot but struggle to understand if it’s boiling soup, making coﬀee, or just

sitting there.

Now, enter multimodal AI. Instead of just looking at the picture, it also reads the recipe text on the counter,

listens to the sizzling sounds, and maybe even detects the smell of spices in the air. It combines information

from the image, text, sound, and smell to get a fuller picture. This way, it can tell not only that there’s a pot on

the stove but also what’s cooking and how far along it is.

In simple terms, while traditional AI might focus on just one piece of the puzzle, like the picture, multimodal

AI brings together information from various sources, mimicking how we use multiple senses to understand our

surroundings. It’s like teaching AI to be more versatile and smarter, just like our own clever brains.

In this article, we explore the interesting world of multimodal NLP. We’ll see how AI tries to understand

things the way we do, using many senses at once.

2.1 What is Multimodal NLP?

Multimodal Natural Language Processing (NLP) is a ﬁeld in artiﬁcial intelligence that focuses on under-

standing and processing language in conjunction with other types of information, or modalities, such as images,

audio, video, and more(Figure 1). In traditional NLP, the emphasis is mainly on working with text data, but

multimodal NLP broadens the scope to incorporate multiple sources of information. The goal of multimodal

NLP is to enable machines to comprehend and generate human language while considering the context provided

by diﬀerent modalities. This approach reﬂects the way humans naturally interact with the world, using various

senses simultaneously.

2.2 Working of Multimodal AI

(image courtesy:https://www.kdnuggets.com/2023/03/multimodal-models-explained.html )

Figure 1: Multimodal learning

(image courtesy:https://www.analyticsvidhya.com/blog/2023/10/exploring-the-advanced-multi-modal-generative-ai/)

Figure 2: Modulus of multimodal Al

For instance, in a multimodal NLP system, understanding a description of a photo might involve not only

analyzing the text but also interpreting the visual content of the image. Similarly, when generating a description,

the system could produce both text and associated visual or auditory information.

The applications of multimodal NLP are diverse and include areas such as image captioning, video

analysis, interactive voice systems, and more. By combining language with other modalities, multimodal NLP

aims to create more robust and human-like interactions between machines and users, capturing the richness of

information available in diﬀerent forms.

2.2 Working of Multimodal AI

Multimodal AI is crafted to understand and generate content across diverse data modes, including text,

images, and audio. This capability is facilitated through three essential modules: the Input Module, Fusion

Module, and Output Module (Figure 2) . Let’s explore these modules to see how Multimodal AI does its job.

2.3 Input Module

The input module in multimodal NLP acts as the eyes and ears of the artiﬁcial intelligence system. It’s the

starting point where the machine takes in information from various forms like written text, images, and audio

recordings. Just as our senses provide information about the world around us, the input module enables the

system to capture a broad range of data, setting the stage for a more comprehensive understanding.

For example, when analyzing a scene with text, images, and spoken words, the input module processes the

written information, recognizes visual elements in the images, and comprehends the spoken words. By handling

multiple modalities, the input module ensures that the AI system can grasp information in a way similar to how

humans use their senses to perceive the world.

2.4 Fusion Module

If the Input Module is like the senses, the Fusion Module is like the brain—it combines and processes

information from diﬀerent sources.Once the information is gathered from diﬀerent modalities, the fusion module

takes center stage. This crucial component is responsible for integrating the diverse data streams into a cohesive

representation.

Imagine you’re watching a video that includes spoken words, text on the screen, and images. The Fusion

Module takes all this information from the Input Module and blends it together. It might synchronize what’s

being said in the audio with the text on the screen and the visuals in the images.

There are diﬀerent approaches to this fusion process.Three prominent techniques come into play here are:

2.4.1 Early Fusion:

In early fusion, information from diﬀerent sources is combined right at the beginning of the system.

Imagine you have a video with spoken words, text on the screen, and images. In early fusion, these diﬀerent

types of information are mixed together right from the start. For example, in our video scenario, early fusion

might involve merging the spoken words, text, and visual information at the initial stages of processing. This

combined information forms a uniﬁed representation that the system then uses to understand the content more

comprehensively. Early fusion is like cooking all the ingredients together from the beginning of a recipe.

It allows the system to consider the connections and relationships between diﬀerent modalities right away,

providing a holistic view of the data.

2.4.2 Alignment-Based Fusion

Alignment-based fusion is a technique in multimodal processing where the system ensures that information

from diﬀerent sources corresponds or aligns correctly. Suppose, you have a multimedia scenario with spoken

words, text, and images. In alignment-based fusion, the system pays special attention to aligning or synchronizing

these diﬀerent modalities to ensure they match up accurately.

For example, if someone in a video is talking about a speciﬁc scene, the alignment-based fusion process

would synchronize the spoken words with the corresponding text on the screen and the relevant visuals in the

images. It’s about making sure that the information from each modality lines up in a way that enhances the

overall understanding. Alignment-based fusion is like carefully arranging pieces of a puzzle to create a complete

picture. By ensuring proper alignment, the system can create a cohesive and accurate representation of the

multimodal data, allowing for a more nuanced and comprehensive understanding.

2.5 Output Module

2.4.3 Late Fusion

Late fusion is a technique in multimodal processing where information from diﬀerent sources is combined

after each source has been processed individually. It’s like combining diﬀerent ﬂavors after each one has been

fully developed. Here’s how it works: Consider a video with spoken words, text on the screen, and images. In

late fusion, each type of information is processed separately ﬁrst. For example, the spoken words are understood

on their own, the text is analyzed independently, and the images are interpreted individually. After these

modalities have been fully processed, the system then combines the information. In our video scenario, late

fusion might involve integrating the fully processed spoken words, text, and visual information. This approach

allows the system to have a more detailed understanding of each modality before blending them together. Late

fusion is like adding distinct ingredients to a recipe one by one and then combining them at the end. By

processing each source independently before fusion, late fusion aims to capture the unique characteristics of

each modality, resulting in a richer and more nuanced representation of the overall data.

To sum it up, the Fusion Module is like the brain of the system. It mixes information from diﬀerent sources,

making sure everything works together smoothly. Whether it combines things early on, lines them up perfectly,

or blends them late in the process, this module helps the system understand the world in a more detailed and

human-like way. It’s like a conductor bringing together diﬀerent instruments to create a beautiful tune, and

as technology gets better, improving how fusion works will make AI even smarter and more helpful in many

diﬀerent areas.

2.5 Output Module

The Output Module in multimodal processing is like the voice or action of the artiﬁcial intelligence system.

Once the input data has been gathered and fused together, the Output Module steps in to generate meaningful

responses or actions. It takes the uniﬁed information and translates it into something understandable or useful.

For example, if you’re interacting with a virtual assistant, the Output Module might generate a text response to

your question, select an appropriate image, or even produce spoken words. It’s the part that communicates the

system’s understanding or decision based on the processed information. The Output Module is crucial because

it turns the combined and processed data into a form that users can comprehend or interact with. It’s like the AI

system expressing itself in a way that makes sense to us, completing the cycle of understanding and interaction.

2.6 Conclusion

In conclusion, the fascinating world of Multimodal Natural Language Processing (NLP) opens new frontiers

in artiﬁcial intelligence, bringing us closer to machines that understand and interact with us in ways that mirror

our own multisensory experience. Through the input module, these systems capture information from diverse

sources like text, images, and audio, resembling our own way of perceiving the world through diﬀerent senses.

The Fusion Module acts as the brain, skillfully blending these varied modalities through approaches like early

fusion, alignment-based fusion, and late fusion. This integration process allows AI systems to understand

small details and show a complete understanding of the information, kind of like how our brains handle details

from various places. Finally, the Output Module transforms this rich understanding into meaningful responses

or actions, completing the cycle of interaction. As technology continues to advance, reﬁning multimodal

NLP techniques holds immense potential for creating more intelligent, versatile, and human-like AI systems.

In essence, the journey into multimodal NLP represents a signiﬁcant stride toward machines that not only

2.6 Conclusion

understand but also engage with us in ways that align more closely with our own complex and dynamic ways of

communication.

References

Multimodal Deep Learning

Multimodal Deep Learning: Deﬁnition, Examples, Applications

A simple guide to multimodal machine learning

Exploring the Advanced Multi-Modal Generative AI

Multimodal Models Explained

Introduction to Multimodal Deep Learning

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.

Integrated Gradients: A Model

Interpretability Technique for Machine

Learning

by Linn Abraham

airis4D, Vol.1, No.12, 2023

www.airis4d.com

3.1 Introduction

In a previous article we saw why interpretable machine learning is important and a survey of the diﬀerent

types of interpretability techniques. In this article we shall look into one speciﬁc technique called Integrated

Gradients [Sundararajan and Taly(2018)]. What is a problem that such techniques try to solve? Suppose we

have a model that has been trained to classify images correctly into one among several classes. Now given an

input image, the model might correctly predict it’s class or it might fail to. If it correctly predicts the class, we

could ask which of the pixels or group of pixels contributed the most to the model’s prediction. This is where

techniques such as Integrated Gradients come to play.

3.2 Applications

The integrated gradients technique can be applied to various types of neural networks. We show the

application to two image models.

3.2.1 Object Recognition Network

An object recognition network that uses a GoogleNet architecture and trained using data from ImageNet

challenge. A black image is used as a baseline and pixel importance is studied using the integrated gradients

technique. The gradients are computed for the output of the highest scoring class with respect to the input pixels

in each image. The results of the study are visualized in Figure. 1.

3.2.2 Diabetic Retinopathy Prediction

Diabetic Retinopathy is a condition that aﬀects the eyesight in patients with diabetes. A deep neural

network has been used to predict the severity grade of the disease using retinal fundus images. The model has

achieved good results on various validation datasets.

3.2 Applications

Figure 1: Comparing integrated gradients with gradients at the image. Left-to-right: original input

image, label and softmax score for the highest scoring class, visualization of integrated gradients, visualization

of gradients*image. Notice that the visualizations obtained from integrated gradients are better at reﬂecting

distinctive features of the image.

3.3 Need for a Baseline

Trust in the networks predictions are important in areas such as health care and pixel attribution techniques

such as integrated gradients can help towards the goal. Figure 2 shows a visualization of integrated gradients

applied to the retinal fundus images. The aggregated gradients are overlayed on a grayscale version of the

image with positive attribution in the green color channel and negative attributions in the red color channel.

The resulting visualization shows a localization of the integrated gradient towards pixels that seem to be actual

lesions in the retina. Also the periphery of the lesion receives positive attribution whereas the interior of the

parts receive negative attribution. Showing that the network focuses on the boundary of the lesions.

Figure 2: Attribution for Diabetic Retinopathy grade prediction from a retinal fundus image. The original

image is show on the left, and the attributions (overlayed on the original image in gray scaee) is shown on the

right. On the original image we annotate lesions visible to a human, and conﬁrm that the attributions indeed

point to them.

3.3 Need for a Baseline

It is common to reﬂect back on a life choice that you made and try to imagine how diﬀerent your life

would have ended up, had your choice been a diﬀerent one. Such thinking is what is called as a counterfactual

thinking. It is even said that the ability to think in counterfactual might be a uniquely human trait. When we

talk about the importance of a feature or pixel, we have in mind a situation where that feature or pixel is absent

which might lead to a better or worse oﬀ situation. A baseline image is useful when one needs to bring in

the concept of missingness of a particular feature. Note that when your input is an image, the features are the

individual pixel values in the image. If you remove a pixel from an image it might lead to several diﬃculties

like changing the shape of your data. Instead if you replace the actual pixel value with zero, you get a black

pixel. Then a pure black image can be considered as a baseline image for all the individual pixels in your input.

But surely such a black image is one among several baselines that can be considered. The impact of diﬀerent

baselines other than a pure black image on attribution methods such as the integrated gradients are covered in

this article [Sturmfels et al.(2020)Sturmfels, Lundberg, and Lee].

3.4 Why not just use Gradients?

Since the gradient of a function represents the direction of maximum increase of the function. The gradients

directly tell us which pixels have the greatest eﬀect on the output. This was why gradients were often used

as a primitive method for such visualizations. However for neural networks the gradients of input pixels may

have small magnitudes around a sample even if the network depends heavily on those pixels. This is called

saturation and can happen when the output function ﬂattens after a certain magnitude is reached for each pixel.

Imagine an image that is interpolated from a pure black image to the sample image. Intuitively this makes

sense, the pixels values immediately close to a given pixel should not change the output function signiﬁcantly.

Additionally the authors of the main paper argue that gradients break the axiom of sensitivity that is required

for any good attribution technique.

3.5 Inspiration from Game Theory

The inspiration for the Integrated Gradients technique comes from Cooperative Game Theory, especially

the Aumann-Shapley cost-sharing method [Aumann and Shapley(1974)]. In games we have sets of participants

which may be called a coalition. The value of a participant or group of participant is computed by understanding

how much the value of the game is increased when the group of participants are added to any given coalition.

3.6 Evaluation of Attribution Techniques

How can we evaluate an attribution technique itself such as the Integrated Gradients? The authors of the

Integrated Gradients paper mention that every such method exisiting in literature had an issue. It could not

distinguish between artefacts that come from perturbing the data, a poor model and a poor attribution technique.

This was one of the motivations for them to come up with an axiomatic way to deﬁne a good attribution method.

3.7 Integrated Gradients

We can consider many paths of interpolating between a baseline and the given input image. Integrated

gradients aggregate gradients along the images that fall on the straight line between the baseline and the input.

The deﬁnition is shown in Figure. 3. Here α is the interpolation coeﬃcient between 0 and 1 which is used to

Figure 3: Deﬁnition of integrated gradients

interpolate between the baseline and the given input image.x

is the given input image and x

′

is the baseline

image. There is still a lot of progress to be made in this area of path attibution methods and interpretablility

methods in general. Which leaves a lot of opportunities for young researchers. This is speciﬁcally useful in

research where the end goal is not just to develop an AI model that can make predictions but also help in ﬁlling

the gaps in theoretical understanding of various phenomenon.

REFERENCES

References

[Sundararajan and Taly(2018)] Mukund Sundararajan and Ankur Taly. A Note about: Local Explanation

Methods for Deep Neural Networks lack Sensitivity to Parameter Values, June 2018.

[Aumann and Shapley(1974)] Robert J. Aumann and Lloyd S. Shapley. Values of Non-Atomic Games. A Rand

Corporation Research Study. Princeton University Press, Princeton, N.J, 1974. ISBN 978-0-691-08103-8.

[Sturmfels et al.(2020)Sturmfels, Lundberg, and Lee] Pascal Sturmfels, Scott Lundberg, and Su-In Lee. Visu-

alizing the Impact of Feature Attribution Baselines. Distill, 5(1):10.23915/distill.00022, January 2020. ISSN

2476-0757. doi10.23915/distill.00022.

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.

Part II

Astronomy and Astrophysics

Black Hole Stories – 5

Some Concepts in Einstein’s General Theory

of Relativity

by Ajit Kembhavi

airis4D, Vol.1, No.12, 2023

www.airis4d.com

1.1 Introduction

In black hole stories-4 we considered the motion or particles around a gravitating object in the framework

of Newton’s theory of gravitation. In this story we will describe some fundamental concepts of general relativity.

These are needed to understand the motion of particles and light rays around black holes, which we will describe

in the next story.

1.2 The Schwarzschild Metric

In the previous story we described the motion of a particle, with small mass m, around a much more

massive particle of mass M. This is like the motion of a planet around the Sun. Both particles were considered

to be point particles, or bodies whose size is very small compared to the distance between them. The more

massive particle can be treated as being at rest, with the smaller mass particle moving in the gravitational ﬁeld

of the larger mass.

We will now consider the same situation, but now from the point of view of the general theory. The massive

particle is the source of the gravitational ﬁeld. But unlike in Newton’s theory, gravitation is no longer a force in

general relativity. The eﬀect of M is to produce a curvature in the space-time structure around M, because of

which the trajectories of particles are diﬀerent from what they would be in the absence of the gravitating mass

M. Our aim is to understand the nature of these trajectories and how they compare with the Newtonian situation.

In Newton’s theory, we considered the classical equations of motion of particles in the gravitational ﬁeld of

M. We used the conservation of energy and angular momentum to analyse the motion, without actually solving

any equations. In general relativity we have to take into account the curved space-time structure. For that, we

have to solve Einstein’s equations to obtain a quantity known as the metric of space-time. The space-time around

the single point mass M we are considering is spherically symmetrical. The eﬀect of M then depends only on

how far the particle is from M, and not on the angular position relative to M. The ﬁrst ever exact solution of

Einstein’s equations was for this spherically symmetric case; it was discovered by Karl Schwarzschild in 1916,

1.3 Geodesics

one year after Einstein announced his equations. The solution provides us the metric of the space-time, from

which all needed information about the space-time can be derived. We will describe below how given such a

space-time, the nature of particle orbits in it can be determined.

In classical mechanics, a spherically symmetric situation is best described in the spherical coordinates r,

θ, φ. Here r is the radial coordinate which measures the separation from the origin at which r=0, and θ, φ are

the usual angular coordinates. In this case, the radial coordinate r indicates distance from the centre. A small

change dθ in the angular coordinate corresponds to a displacement rdθ, while a change dφ in the φ coordinate

corresponds to a displacement rsinθdφ. The area of a sphere of radius r is 4πr

. A spherical coordinate system

is shown in Figure 1.

Figure 1: A spherical polar coordinate system. Diagram Courtesy Knowino Contributors

The situation is somewhat diﬀerent for the Schwarzschild case, where again spherical polar coordinates r,

θ, φ are used. In this case a consequence of the space-time curvature is that r is no longer the distance from the

origin. And yet displacements corresponding to small changes dθ and dφ are again rdθ and rsinθdφ. How is

that possible? The reason is that the displacements correspond to small changes around a given point r, θ, φ,

while ﬁnite values of r are diﬀerent from the actual distance from the origin. The distance, if needed, can be

calculated using the metric. Interestingly, the area of a sphere corresponding to r is again 4πr

1.3 Geodesics

According to Newton’s ﬁrst law of motion, a free particle, that is one which has no force acting on it,

moves in a straight line with a constant velocity. This is also true in Einstein’s special theory of relativity.

When a force is present, there is deviation from a straight line. We have seen in story-4 how particle trajectories

are determined in classical mechanics when a gravitational force is present. The situation is quite diﬀerent in

general relativity, where gravitation is identiﬁed with the structure of space-time, so gravity is always present.

A free particle in general relativity is one which has nothing except gravity aﬀecting its motion. The trajectory

of such a particle is known as a geodesic.

The shape of a geodesic is quite diﬀerent from that of straight line. As an example, let us consider the

surface of a sphere and any two points A and B on it, as shown in Figure 2. The shortest distance between the

two points, while remaining on the surface of the sphere, is the segment of a great circle which passes through

the two points. A great circle is a circle on the spherical surface which has the centre of the sphere as its own

centre. If we considered the situation in three dimensions, we could have drawn a straight line joining A and

B. But if we are restricted to remining on the curved surface, then the shortest distance is given by the segment

AB. In general relativity we are inescapably in a curved space-time. So we have to use geodesics.

Figure 2: A spherical surface with a great circle passing through points A and B. The great circle is a geodesic

for the spherical surface. The segment of the geodesic joining A and B is the shortest distance between them.

Image (Courtesy Vikas Mittal http://dx.doi.org/10.48550/arXiv.2209.04810 )

1.4 Types of Geodesics

How does one deﬁne a geodesic? In the type of geometry, known as Riemannian geometry, which

correspond to the space-time of general relativity, a geodesic can be deﬁned as the curve of shortest distance

between two given points. The distance here is a 4-dimensional space-time analogue of the simple three

dimensional distance that we are used to. The distance can be determined once the metric of the space-time is

known. We will consider in the next story the geodesics of the Schwarzschild metric which correspond to the

trajectories of planets around the Sun.

1.4 Types of Geodesics

In special and general relativity, no particle can travel faster than light. This fundamental fact allows us to

deﬁne three types of geodesics: timelike, null and spacelike. Time like geodesics are those which correspond to

the trajectories of particles which move with speed less than the speed of light, like ordinary particles do. Null

geodesics correspond to particles which move with the speed of light. These necessarily have zero rest mass,

like photons which are particles of light. If neutrinos have zero rest mass they will move on null geodesics too,

but if they are massive they will move along time like geodesics. Spacelike geodesics correspond to trajectories

of particles which move faster than the speed of light. No such particles have so far been found, but there have

been some discussions about particles called tachyons which move faster than light.

In 4-dimensional space-time, the location of any event has 4 coordinates, three space coordinates describing

where an event has occurred, and a fourth time coordinate describing when it has occurred. For example a

ball could be thrown from a point on surface of the Earth at a time t

, which constitutes an event A. The ball

could be caught at another point at time t

later than t

, which constitutes event B. The events A and B have a

timelike separation since the ball travels from A to B at a speed which is evidently less than the speed of light.

If instead, A and B consisted of a beam of light being emitted at A and received at B, then the events A and B

have null separation. In both these cases event A can inﬂuence B, since a signal can be passed from A to B. But

suppose the distance between the events is such that not even a beam of light can travel from A to be in the time

-t

Then the two events are said to have a spacelike separation. In such a case event A cannot inﬂuence event

1.5 Conserved Quantities

In Story-4, we saw how in the classical case, the study of particle trajectories is signiﬁcantly simpliﬁed

by using the conserved quantities of energy E and angular momentum L. With these quantities we could get

important insights into the trajectories without having to solve the diﬀerential equations of motion. We would

like to take a similar approach in general relativity, but for that we need to be able to deﬁne conserved quantities.

Because of the complexity of space-time curvature, it is not obvious how quantities like energy and angular

momentum are to be deﬁned in the theory. But then Emmy Noether’s theorem comes to our rescue.

In classical mechanics, it is possible to deﬁne energy and angular momentum from elementary consider-

ations. But we described another route using Noether’s theorem, where a correspondence is set up between

conserved quantities and symmetries of the situation. Then energy conservation follows because the gravi-

tational ﬁeld of a point particle is independent of time, and the conservations of angular momentum follows

from the spherical symmetry. It turns out that Noether’s theorem can be used in general relativity too. When

a symmetry is present, there is a corresponding quantity called a Killing vector. Each Killing vector allows a

conserved quantity to be deﬁned.

1.5 Conserved Quantities

The Schwarzschild solution is independent of time, and has spherical symmetry, which leads to two Killing

vectors and two corresponding conserved quantities which we can call E and L. When the particle trajectory is

far from the centre, where the gravitational ﬁeld is weak, the eﬀect of space-time curvature is not important.

There E and L can be interpreted as energy and angular momentum respectively. When the particle is close

to the centre and space-time curvature is signiﬁcant, E and L continue to be interpreted as energy and angular

momentum, in analogy to what happens at a great distance. But more than the meaning of these terms, the fact

that they are conserved is important.

We will see in the next story how the above fundamental considerations can be used to study particle

trajectories and how interesting the results are.

[Will continue..]

About the Author

Professor Ajit Kembhavi is an emeritus Professor at Inter University Centre for Astronomy and

Astrophysics and is also the Principal Investigator of the Pune Knowledge Cluster. He was the former director

of Inter University Centre for Astronomy and Astrophysics (IUCAA), Pune, and the International Astronomical

Union vice president. In collaboration with IUCAA, he pioneered astronomy outreach activities from the late

80s to promote astronomy research in Indian universities. The Speak with an Astronomer monthly interactive

program to answer questions based on his article will allow young enthusiasts to gain profound knowledge about

the topic.

Exploring the Enigmatic World of Be/X-ray

Binaries

by Sindhu G

airis4D, Vol.1, No.12, 2023

www.airis4d.com

2.1 Introduction

X-ray binaries, illustrated in Fig: 1, represent binary star systems where one of the stars, often a compact

object like a neutron star or a black hole, attracts and accumulates matter from its companion star. Among

the classiﬁcations of X-ray binaries, high-mass X-ray binaries hold a signiﬁcant position. A high mass X-ray

binary (HMXB) is a speciﬁc kind of binary star system featuring a high-mass star (typically an O or B-type

star) paired with a compact object such as a neutron star or a black hole. These binary systems are identiﬁed

by their intense X-ray emissions, a consequence of the interaction between the massive star and the compact

object. Be/X-ray binaries emerge as a notable subtype within the category of high-mass X-ray binary systems.

Be/X-ray binaries are the most abundant type of massive X-ray binary systems in the Galaxy. This article delves

into the fascinating realm of Be/X-ray binaries, exploring the components, characteristics, and the scientiﬁc

signiﬁcance of these captivating cosmic partnerships.

2.2 Understanding Be/X-ray Binaries

Be/X-ray binaries(Fig: 3) are a class of high-mass X-ray binaries.

2.2.1 The Components

Be/X-ray binaries consist of a Be star and a neutron star. Be stars typically belong to the O or B spectral

classes and are renowned for their strong stellar winds and the formation of circumstellar disks. Be stars,

characterised by their signiﬁcant mass and high temperature, undergo rapid rotation. Be stars are known for

their emission lines arising from circumstellar disks composed of expelled material and gas. These circumstellar

disks play a crucial role in the dynamics of Be/X-ray binaries, inﬂuencing mass transfer and shaping the overall

behaviour of the binary system. Predominantly studied emission lines include those of hydrogen, such as the

Balmer and Paschen series, while Be stars may also display emission lines of helium and iron. In comparison to

B stars of the same optical spectral type, Be stars exhibit stronger infrared radiations, leading to the phenomenon

known as infrared excess. The mechanism behind the formation of such a circumstellar disk remains an open

question, generally attributed to the rapid rotation of Be stars. Given the extended semi-major axis of the

2.2 Understanding Be/X-ray Binaries

Figure 1: Artist’s conception of an X-ray binary system.[ Credit: NASA/GSFC ]

Figure 2: Classiﬁcation of X-ray binaries[ Credit: Pablo Reig]

2.2 Understanding Be/X-ray Binaries

Figure 3: Sketch of a Be/X-ray binary system[ Credit: M. Orellana & G.E. Romero ]

orbital path in these sources, the volume of the companion star is notably smaller than that of the Roche lobe.

Consequently, due to the inﬂuence of the eccentric orbit, the compact star typically remains at a considerable

distance from the peripheral disk of the Be star, maintaining a relatively quiescent state for an extended duration.

The secondary star in the system is the neutron star, responsible for generating X-rays. Neutron stars,

originating from the collapse of massive star cores, are exceptionally dense and compact objects. These stars

possess robust magnetic ﬁelds and are capable of rapid rotation.Typically, the neutron star orbits the Be star in a

wide and highly elliptical path. The Be star’s stellar wind gives rise to a disk, usually oriented diﬀerently from

the neutron star’s orbital plane.

2.2.2 X-ray Emission

When the neutron star traverses the Be disk, it rapidly accumulates a substantial amount of gas. The

infalling gas onto the neutron star results in the observation of a luminous ﬂare in hard X-rays. The gravitational

forces at play during the accretion process generate intense temperatures and pressures, causing the release of

X-rays. This phenomenon makes Be/X-ray binaries prominent X-ray sources in the sky. Observations in the

X-ray spectrum allow astronomers to study the properties and behaviour of these systems.

The occurrence of X-ray emission bursts in these sources is attributed to the interaction when the dense

companion star approaches the circumstellar disk of the Be star. During this interaction, the compact star

accretes matter characterised by low rotation speed and high density, a process that can also occur through

Roche lobe outﬂow. At this juncture, the brightness of the X-ray radiation near the hot point of the circumstellar

disk undergoes a signiﬁcant increase, giving rise to bursts that endure for weeks to months, as depicted in Fig:

2.2.3 Transient Astrophysical Phenomena

Be/X-ray binaries are known for their transient behaviour, characterised by unpredictable X-ray outbursts.

The transient astrophysical phenomena in Be/X-ray binaries are integral to understanding the dynamic nature of

these systems. The observed outbursts, accretion processes, and variations during periastron passages contribute

valuable data for researchers seeking to unravel the intricacies of mass transfer, disk dynamics, and the overall

2.3 Observational Techniques

Figure 4: Formation and evolution of compact stellar X-ray sources[ Credit: T. M. Tauris & E. P. J. van den

Heuvel ]

behaviour of Be/X-ray binaries.

2.3 Observational Techniques

Astronomers employ a variety of observational techniques to study Be/X-ray binaries. Observations in

X-rays, optical, IR and radio wavelengths provide a comprehensive understanding of the systems, allowing

scientists to piece together the puzzle of their characteristics and behaviours.

X-ray astronomy plays a pivotal role in the study of Be/X-ray binaries, where X-rays stand as the predominant

emission source. X-ray telescopes become indispensable tools for measuring the X-ray spectrum, ﬂux, and

variability exhibited by Be/X-ray binaries. This data is instrumental in investigating the intricacies of the

accretion process and discerning the properties associated with the compact object.

Optical astronomy employs telescopes to scrutinise the Be star component within BeXRBs. This involves

measuring the star’s spectrum, luminosity, and variations in brightness. Additionally, optical spectroscopy is

utilised to examine the circumstellar disk surrounding the Be star.

The utilisation of radio telescopes enables the detection of radio emissions emanating from BeXRBs.

These emissions result from jets of material expelled from the accretion disk. Radio observations provide a

means to scrutinise the characteristics of both the jets and the accretion disk.

Infrared astronomy employs telescopes to examine the circumstellar disk surrounding the Be star. This

involves measuring the temperature, density, and dust content of the disk. Additionally, infrared observations

are instrumental in studying the mass loss rate from the Be star.

2.4 Importance of the Study of Be High-Mass X-ray Binaries

Be/X-ray binaries (BeXRBs) function as distinctive laboratories that explore extreme conditions in as-

trophysics. Comprising a Be star and a neutron star, these binaries oﬀer unique opportunities to examine the

properties of matter under intense gravitational forces and magnetic ﬁelds.

These systems are known for their high variability across various time scales and wavelengths, providing

valuable insights into the dynamic behaviour of matter within them. The accretion of material from the Be

star’s decretion disk onto the compact object, usually a neutron star, results in bright ﬂares in hard X-rays. This

2.4 Importance of the Study of Be High-Mass X-ray Binaries

process not only sheds light on accretion physics but also contributes to our understanding of mass transfer

mechanisms in BeXRBs.

The neutron star’s tidal force plays a crucial role in truncating the Be star’s decretion disk, causing

equatorial disks in BeXRBs to be smaller and denser compared to those around isolated Be stars. Type I

outbursts, traditionally associated with eccentric binary orbits, have been proposed to occur in nearly circular

orbit Be/X-ray binaries due to the presence of the 3:1 Lindblad resonance within the Be star disk.

BeXRBs also serve as platforms for studying mass determination of the neutron star and the evolution of

its spin period. These investigations contribute to our broader understanding of the properties of neutron stars

and Be stars.

In summary, the signiﬁcance of Be/X-ray binaries extends beyond their role as astronomical phenomena.

They act as crucial laboratories, oﬀering insights into accretion processes, mass transfer dynamics, the eﬀects of

tidal forces, and the intricacies of neutron star properties. As dynamic astrophysical systems, Be/X-ray binaries

contribute signiﬁcantly to advancing our understanding of extreme conditions and fundamental processes in the

cosmos.

Various categories of Be X-ray binaries will be explored in the upcoming article.

References:

High-Mass X-ray binary: Classiﬁcation, Formation, and Evolution

Optical/infrared observations unveiling the formation, nature and evolution of High-Mass X-ray Binaries

Be/X-ray binary

A catalogue of high-mass X-ray binaries in the Galaxy: from the INTEGRAL to the Gaia era

Be/X-Ray Binaries

Gamma-Ray Emission from Be/X-ray Binaries

Formation and evolution of compact stellar X-ray sources

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

Exploring the Complexities of Ageing Process

with respect to Biological Age and

Chronological Age

by Geetha Paul

airis4D, Vol.1, No.12, 2023

www.airis4d.com

Image Courtesy: https://bioethics.hms.harvard.edu/journal/legal-age-change

1.1 Introduction

The concept of ageing can be approached through the lens of biological age and chronological age,

both of which are distinct ways to measure a person’s age. Chronological age refers to the amount of time

elapsed from birth to a given date or it is a conventional measure based on the number of years lived, it is an

imperfect indicator of the ageing process. While biological age considers various biological and physiological

developmental factors, such as genetics, lifestyle, nutrition, and comorbidities. The unravelling of the biology

of senescence has enabled the identiﬁcation of molecular markers of biological age, providing insights into

the cellular level of ageing. This article provides a conceptual overview of biological age, emphasising the

limitations of chronological age and the signiﬁcance of biological age biomarkers in understanding the ageing

process. The distinction between chronological and biological age is crucial, as it has implications for health-

related outcomes and interventions aimed at extending healthspan and lifespan. As the mysteries of ageing are

being demystiﬁed, the potential to alter the course of someone’s lifespan and slow the pace of ageing is becoming

conceivable. This article highlights the importance of exploring the complexities of the ageing process through

the lens of biological and chronological age, shedding light on the potential for interventions and advancements

in understanding longevity and healthy ageing.

1.2 Telomere Shortening

Biological age is a comprehensive concept that is not only related to lifespan but also other health-related

outcomes, including quality of life. Therefore, biological age provides a more accurate assessment of the ageing

process compared to chronological age. Recent developments in epigenetics suggest that our epigenome might

be the key to understanding and calculating human biological ageing. This implies that it may soon be possible

to accurately calculate a person’s biological age. Currently, legal age always corresponds with chronological

age, but if a person’s biological age turns out to be diﬀerent, it raises the question of whether legal age should be

based on biological rather than chronological age. This distinction could have implications for how well people

function, as biological age is a better indication of a person’s ability and well-functioning. The potential to

accurately calculate biological age based on epigenetic changes may lead to discussions about legal age change

and its alignment with biological rather than chronological age. The conventional view of ageing is that it is

a linear process of physical and cognitive decline that occurs over time as one progresses from adulthood into

senescence. The Ageing process can be deﬁned by a progressive loss of complexity within the dynamics of

physiologic outputs (Lipsitz and Goldberger, 1992).

The ageing process is a multifactorial phenomenon that commences at the cellular level, leading to a gradual

decline in the body’s larger systems. Various theories, such as oxidative stress, mitochondrial dysfunction, cell

senescence, and telomere shortening, have been proposed to explain the ageing process. Genetic factors,

particularly telomere length, have been identiﬁed as potential ageing biomarkers.

1.2 Telomere Shortening

Telomeres, which are protective structures at the end of chromosomes, undergo shortening with age,

and this shortening rate may indicate the pace of ageing. Additionally, T-cell DNA rearrangement and DNA

methylation have been recognized as ageing biomarkers.

Image Courtesy: https://docs.google.com/document/d/1xRJb4w4FeuAMCUbSC9t4dSdr4MWeu5XWL-bAilDVVk8/edit

Figure 1: A process termed replicative senescence. (Hayﬂick and Moorhead, 1961; PMID: 13905658). Due

to signals derived from the trimming of the ends of chromosomes– called telomeres– that occurs with each

successive round of DNA duplication prior to cell division [(Bodnar AG, 1998; PMID 9454332),(Kuilman T

2010, PMID 21078816)

1.3 Epigenetic Regulation

Epigenetic regulation is key to the normal development of an organism because it allows for modulating gene

expression levels through the addition of chemical modiﬁcations to the DNA and its associated histone proteins,

which are referred to as epigenetic marks. Cells of a multicellular organism are genetically homogeneous

but structurally and functionally heterogeneous owing to the diﬀerential expression of genes. Many of these

diﬀerences in gene expression arise during development and are subsequently retained through mitosis. Stable

alterations of this kind are said to be ’epigenetic’, because they are heritable in the short term but do not involve

mutations of the DNA itself. Research over the past few years has focused on two molecular mechanisms

that mediate epigenetic phenomena,DNA methylation and histone modiﬁcations. Epigenetic eﬀects by means

of DNA methylation have an important role in development but can also arise stochastically as animals age.

Identiﬁcation of proteins that mediate these eﬀects has provided insight into this complex process and diseases

that occur when it is perturbed. External inﬂuences on epigenetic processes are seen in the eﬀects of diet on

long-term diseases such as cancer. Thus, epigenetic mechanisms seem to allow an organism to respond to

the environment through changes in gene expression. The extent to which environmental eﬀects can provoke

epigenetic responses represents an exciting area of future research.

Image Courtesy: https://www.mdpi.com/2073-4409/10/5/1074

Figure 2: Schematic of the mechanisms of epigenetic regulation. DNA methylation, histone modiﬁcations and

chromatin remodelling represent three diﬀerent kinds of epigenetic mechanisms. Main players in the histone

modiﬁcation machinery are depicted in the inset. Proteins that covalently attach chemical groups to the histone

tails are termed writers, whereas the so-called readers can recognize and bind to histone modiﬁcations. Enzymes

that remove histone marks are termed erasers. DNA methylation is found across diﬀerent genomic elements. For

example, CpG islands, which are often found in the proximity of promoters, are usually depleted of methylation,

whereas gene bodies are heavily methylated. The chromatin structure is controlled by chromatin remodeler

complexes that use the hydrolysis of ATP to mediate the packaging of the chromatin.

1.4 DNA Methylation and Histone Modiﬁcations

DNA methylation, an epigenetic modiﬁcation, has been found to play a signiﬁcant role in modulating gene

expression and is associated with age-related pathologies such as cancer, osteoarthritis, and neurodegeneration.

Furthermore, DNA methylation biomarkers have been shown to determine the biological age of tissues across

the human lifespan and can estimate biological ageing rates. The potential of DNA methylation as a biological

biomarker for age assessment is being explored, and it is considered a promising avenue for understanding

1.5 CpG Islands

and addressing age-related pathologies. T-cell DNA rearrangement and DNA methylation play crucial roles

in T-cell development, diﬀerentiation, and function. DNA methylation and its enzymatic regulators direct

the development and diﬀerentiation of CD4+ and CD8+ T-cells. In T and B lymphocytes, demethylation at

speciﬁc loci has been associated with the transcription and rearrangement of immunoglobulin and TCR genes,

inﬂuencing cell function and survival. The epigenetic processes, including DNA methylation, regulate the

primary rearrangement of a single allele, impacting T-cell receptor gene assembly. These ﬁndings highlight

the signiﬁcance of DNA methylation and T-cell DNA rearrangement in T-cell development, function, and

ageing-related pathologies.

DNA methylation is an epigenetic mechanism that involves the addition of a methyl group to the 5th

position of a cytosine in a CpG dinucleotide context. It is essential for mammalian development and gene

regulation. Key aspects of DNA methylation include: Gene Silencing: DNA methylation is mainly associated

with gene silencing, as it prevents proteins from binding to the methylated DNA. In the mammalian genome,

70-80% of CpG dinucleotides are methylated

1.5 CpG Islands

: CpG-rich sequences in promoter regions, referred to as CpG islands, usually lack DNA methylation. These

islands are crucial for stable gene expression and the regulation of imprinting. Gene Body Methylation: Methy-

lation of the gene body is associated with higher gene expression in dividing cells. DNA Methyltransferases

(DNMTs): These enzymes transfer a methyl group to the C5 position of cytosine, forming 5-methylcytosine.

The DNMT family includes DNMT1, DNMT2, and DNMT3. Demethylation: DNA methylation can be ac-

tively removed by members of the ten-eleven translocation (TET) enzyme family. These erasers mediate DNA

demethylation by oxidising 5-methylcytosine to 5-hydroxy-methylcytosine.

Image Courtesy: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10488447)

Figure 3: The eﬀects of advancing age on the vasculature. Vascular ageing is characterised by arterial stiﬀness

and calciﬁcation. Arterial stiﬀness can induce an increase in vascular resistance and organ dysfunction. Arterial

calciﬁcation induced thrombotic events. A feed-forward loop exists between vascular ageing, hypertension, and

arterial stiﬀness.

1.6 Biomarkers of Biological Vascular Aging

(Image Courtesy: (https://academic.oup.com/lifemedi/advance-article/doi/10.1093/lifemedi/lnad033/7255917)

Figure 4: Shows the Biomarkers of Biological Vascular Aging

1.6 Biomarkers of Biological Vascular Aging

The consensus statement by the Aging Biomarker Consortium provides a comprehensive assessment

of biomarkers associated with vascular ageing (VA) and classiﬁes them into three dimensions: functional,

structural, and humoral. The expert panel recommends clinically relevant VA biomarkers within each dimension,

including those reﬂecting vascular stiﬀness, endothelial function, vascular structure, microvascular structure,

distribution, and proinﬂammatory factors. The concept of biological age has been introduced as an integrated

measure reﬂecting the individualised ageing pace, emphasising the importance of identifying biomarkers that

outperform chronological age as determinants of morbidity and mortality. It proposes six pillars for ageing

biomarkers, including physiological characteristics, medical imaging, histological features, cellular alterations,

molecular changes, and secretory factors, to fulﬁl the requisites of assessing age-related changes, tracking the

physiological ageing process, and predicting the transition into a pathological status.

Some of the key biomarkers include: Pulse Wave Velocity (PWV)The: The most common measure of

arterial stiﬀness, PWV correlates with chronological age and is associated with an increased risk of CVD and

all-cause mortality Carotid Intima-Media Thickness (IMT): A measure of subclinical atherosclerosis, carotid

IMT increases with age and is associated with CVD morbidity and mortality. Coronary Artery Calcium Score

(CACS): A noninvasive measure of atherosclerosis, CACS is a powerful predictor of incident CVD and all-

cause mortality. Flow-Mediated Dilation (FMD): A measure of endothelial dysfunction, FMD decreases during

ageing and is an independent predictor of CVD outcomes. hs-CRP, Folic Acid, Homocysteine, and Fibrinogen:

These biomarkers have been associated with early vascular damage and are used to assess CVD risk.

Vascular ageing is characterised by the deposition of calcium phosphate crystals in both the arterial intima

(typically related to atherosclerosis) and the media (called M’nckeberg sclerosis). Both types of calciﬁcation

often develop in parallel and are not always easily distinguished by imaging techniques. Computed tomography

is the gold standard technique for quantiﬁcation of coronary artery calcium score (CACS). Even though CACS

can be inﬂuenced by the presence of medial calciﬁcation, it is typically used as a surrogate marker for the extent

of atherosclerosis because it correlates well with coronary plaque burden. CACS correlates with chronological

age and is a powerful predictor of incident CVD and all-cause mortality.

1.7 Conclusion

(Image Courtesy: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10488447/)

Figure 5: Biomarkers for Vascular Aging Assessment

Vascular ageing is also characterised by atherosclerosis, which consists of lipid-rich plaque buildup in

the intima that can lead to acute myocardial infarction or stroke. Endothelial dysfunction, a major driver of

atherogenesis, can be measured by ultrasound as ﬂow-mediated dilation, and ﬂow-mediated dilation decreases

during ageing and is an independent predictor of CVD outcomes.

These biomarkers can provide valuable insights into the vascular ageing process and help identify individ-

uals at high risk of developing cardiovascular disease. However, further research is needed to establish their

clinical relevance and improve their accuracy in predicting and managing vascular ageing-related conditions.

1.7 Conclusion

Understanding the complexities of the ageing process through the lens of biological and chronological age,

as well as vascular age, sheds light on the potential for interventions and advancements in understanding longevity

and healthy ageing. The distinction between biological and chronological age, along with the identiﬁcation

of biomarkers for vascular ageing, has implications for various aspects of life, including legal age change

and healthcare decisions. Recent developments in epigenetics suggest that the epigenome might be the key

to understanding and calculating human biological ageing, and advancements in understanding longevity and

healthy ageing are being made. The identiﬁcation and utilisation of biomarkers for vascular ageing can improve

early detection of individuals at high risk of developing cardiovascular disease, and further research is needed to

establish their clinical relevance and improve their accuracy in predicting and managing vascular ageing-related

conditions.

References:

https://www.nature.com/articles/ng1089z

https://www.mdpi.com/2073-4409/10/5/1074

https://academic.oup.com/lifemedi/advance-article/doi/10.1093/lifemedi/lnad033/7255917

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8712891/

1.7 Conclusion

https://www.jacc.org/doi/10.1016/j.jacc.2019.11.062

Lipsitz LA, Goldberger AL. Loss of ‘complexity’ and ageing. Potential applications of fractals and chaos

theory to senescence. JAMA. 1992;267:1806–1809. [PubMed]

[Google Scholar]

Salih A, Nichols T, Szabo L, Petersen SE, Raisi-Estabragh Z. Conceptual Overview of Biological Age

Estimation. Ageing Dis. 2023 June 1;14(3):583-588. doi: 10.14336/AD.2022.1107.PMID:37191413; https:

//www.ncbi.nlm.nih.gov/pmc/articles/PMC10187689/

Ahadi, Sara, Wenyu Zhou, Sophia Miryam Schssler-Fiorenza Rose, M. Rexa Sailani, K vin Contrepois,

Monkia Avina, Melanie Ashland, Anne Brunet, and Michael Snyder. Personal ageing markers and ageo-

types revealed by deep longitudinal proﬁling. Nature Medicine 26 (2020): 83-90. https://doi.org/10.1038/

s41591-019-0719-5.

Sillanpaa, Elina, Miina Ollikainen, Jaakko Kaprio, Xiaoling Wang, Tuija Leskinen, Urho M. Kujala, and

Timo Trmkangas. Leisure-time physical activity and DNA methylation age—a twin study. Clinical Epigenetics

11, 1 (2019): 12. https://doi.org/10.1186/s13148-019-0613-5

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

Workaholism and Depression

by Ninan Sajeeth Philip

airis4D, Vol.1, No.12, 2023

www.airis4d.com

1.1 The Myths about Workaholism

Motivation is the guiding force of life. The less a person is self-motivated, the higher the chance for them

to go down the valley of depression. Self-motivation should not be mistaken for workaholism. Self-motivation

promotes hard work. The main diﬀerence between a hardworking person and a workaholic lies in the underlying

characteristics and the impact on their personal and professional lives. There are many simple and reliable

ways to distinguish the two. The central element is motivation itself. If your work makes you happier, it is

positive. On the other hand, it is something to be concerned about if it adds to stress and lack of conﬁdence and

self-esteem. A healthy sense of self-worth and enjoyment in completing even small tasks are the most apparent

characteristics of a healthy individual.

In the modern world, where everything counts in reward, people opt for professions they are incapable of -

it could be peer or parental pressure. Though educated, many parents still mislead children from childhood to

achieve what they cannot and push them to the extremes of stress and obsessiveness with the consequence of

1.2 Fundamental Diﬀerences

depression and lack of self-worth.

Flexibility and adaptability to unexpected turnarounds are characteristic of a healthy mind. Failures are the

less-known side of success; nobody has achieved anything without facing failure. Occasional depressions are

also ﬁne. It is a vent to exhaust emotions and to get back to routine. While the time and mode of overcoming

depressive stages vary between individuals, the concern is when it becomes obsessive or prolonged beyond

acceptable periods.

Another noticeable characteristic of workaholics is their withdrawal from social and community life.

They tend to be more self-centric, withdrawing themselves from public spaces and lacking the courage to

face the public. While environmental changes can trigger depression, it is more or less understood to have

genetic links that play a signiﬁcant role in mood disorders and obsessive behaviour. One well-studied and

established association is the genetic link of Obsessive Compulsive Disorder (OCD), which is a pathway to

clinical depression. Clinical depression is the state in which an individual is feeling low and worthless for a

couple of weeks or more, highly irritable and fussy in their relationship with others.

On the other hand, hard workers are active in their public life, ﬁnding it easy to adapt to the situation and

make the best of it by living in the moment. Living in the moment is cultivated and cherished when one learns

to emphasise the importance of seeing opportunities in everything. The diversity in their interests makes them

unique and signiﬁcantly aﬀects their success. While workaholics pretend, hard workers get involved and adapt

to the environment, turning even disasters into opportunities.

1.2 Fundamental Diﬀerences

Health and well-being are yet another distinctive characteristic between hard workers and workaholics.

Workaholics are more likely to experience health issues due to their compulsive behavioural habits. In contrast,

hard workers burn out their excessive energy in activities that maintain the level of dopamine in their brains,

thereby continuing a balanced work-life. The passion and internal drive help them to focus on their goals to move

forward irrespective of the environmental or social campaigns. They may be concerned but are not aﬀected

by how others judge or evaluate them to the extent that workaholics are. They can overcome social pressures

by deciding their goals, knowing the routes, and being prepared to take risks. On the other hand, workaholics

seek social validation and appreciation and push even to complete mundane tasks. In brief, the motivation for a

workaholic is external, while a hard worker’s inspiration is from within.

There are simple methods to identify and keep away from getting trapped in depression. The ﬁrst and

foremost emphasis has to be on introspection. Introspection is to look back at yourself and ask simple questions

like whether what you are doing is stressful, resulting in low self-esteem and anxiety. Stress and hard work

are diﬀerent, and one should see the diﬀerence. Hard work is essential for success and healthy self-esteem. I

remember the principal in college coming to our ﬁrst day in the postgraduation class and saying, “I have nothing

more to oﬀer you but two years of hard work and sweating!” That is a great oﬀer I often share with my students.

Like the old saying “No crown without a cross”, hard work and success are synonyms, and one cannot earn one

without the other. Harworkers have sleepless nights and sweating days, but their passion or internal drive is the

key diﬀerentiating factor. If that is missing, it is time to rethink and turn away. While neurotransmitters like

serotonin cause the state of depression, exposure to stressful life, such as living in an abusive relationship or

prolonged exposure to stress at work, can catalyst the risk of depression.

1.3 Conclusion

Thomas Edison is known for his strong work ethic. He held 1093 patents in his industrial research

laboratory, mostly singlehandedly. Edison had a lot of critics at his time, and while most praised him, many

well-known scientists criticised him for the foolishness that is indicative of his ignorance of basic science.

However, Edison was willing to accept his fallacies and is said to have commented that he had made thousands

of discoveries of things that do not work before identifying what works. This is the attitude of a hard worker.

They see the brighter side of what the workaholic would blame as the cause of their depression and failure.

Reference

[1] 20 Diﬀerences between a workaholic and a hard worker - Managers Orbit https://www.managersorbit.

com/workaholic-vs-hard-worker/

[2] The Diﬀerence Between Hard Worker and Workaholic | Blog - Palmer Group https://www.thepalmergroup.

com/resources/blog/the-ﬁne-line-between-hard-worker-and-workaholic

[3] Are You a Hard Worker or a Workaholic? - Inc. Magazine https://www.inc.com/carolyn-cutrone-the-diﬀerence-between-workaholic-and-hard-worker.

html

[4] How Being a Workaholic Diﬀers from Working Long Hours — and Why That Matters for Your Health -

Harvard Business Review https://hbr.org/2018/03/how-being-a-workaholic-diﬀers-from-working-long-hours-and-why-that-matters-for-your-health

[5] Workaholism facts and statistics: everything you need to know - Clockify https://clockify.me/workaholism-facts

[6] 4 Famous Workaholics (And The Secrets of Their Success) - Lifehack https://www.lifehack.org/articles/

work/4-famous-workaholics-and-the-secrets-of-their-success.html

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.