Cover page
Species Name: Ictinogomphus rapax.
Ictinogomphus rapax,the common clubtail, is a species of dragonfly (https://en.m.wikipedia.org/wiki/Dragonfly)
in the family Gomphidae (https://en.m.wikipedia.org/wiki/Gomphidae) photographed from airis4D campus.
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
i
Editorial
by Fr Dr Abraham Mulamoottil
airis4D, Vol.1, No.10, 2023
www.airis4d.com
Welcome to the 10th edition of airis4D Journal, where we embark on a journey through diverse realms
of scientific inquiry and exploration. This edition brings together a collection of articles that delve into the
frontiers of quantum computing, data analysis, astronomy, biology, and the art of effective presentation.
1. Quantum Leap in Machine Learning: Part 1 In the opening article, we dive into the groundbreaking
intersection of quantum computing and machine learning (QML). The narrative unfolds the evolution of
QML from its early conceptualization in the 2000s to the present day. From Quantum Support Vector
Machines to Deep Quantum Neural Networks, the article presents a landscape where quantum capabilities
redefine the boundaries of data processing. The author, Blesson George, underscores the unique features
of quantum computers—superposition and entanglement—and their pivotal role in overcoming challenges
posed by classical logic-based computing when dealing with quantum phenomena.
2. Unraveling Clustering Methods: Decoding Patterns in Data Jinsu Ann Mathew takes us on a tour of
the intricate world of clustering methods, unraveling the significance of grouping similar data points to
unveil hidden structures. Through a lens focused on an e-commerce platform, the article showcases the
artistry involved in selecting the right clustering technique for different datasets. Mathew’s expertise in
Natural Language Processing and Chemical Informatics lends credibility to the insights shared, making
this exploration into clustering not just informative but profoundly practical.
3. Stellar Symphony: A Celestial Classification Ballet Robin Jacob Roy conducts a celestial symphony in
the exploration of spectral classification, a method that unveils the secrets of stars. From the historical
origins with Angelo Secchi to the detailed classification developed at the Harvard College Observatory,
the article provides a stellar performance. Roy’s expertise in Physics and Machine Learning applications
for Astronomy adds depth to the narrative, emphasizing the crucial role of spectral classification in
understanding the diverse nature of stars and contributing to broader astronomical explorations.
4. Beyond Hubble: Morphological Galaxy Adventures Dr. Sheelu Abraham takes us beyond the iconic
Hubble tuning fork diagram into the intricate world of galaxy classification. Exploring schemes such as
Morgan’s Yerkes System, de Vaucouleurs Classification, Elmegreen’s Classification, and van den Bergh
Classification, the article offers a nuanced perspective. The conclusion resonates with the simplicity
and foundational significance of Hubble’s classification, recognizing the iterative nature of scientific
advancement.
5. Cosmic Partnerships: High-Mass X-ray Binaries Unveiled Sindhu G guides us through the cosmic
partnerships in High-Mass X-ray Binaries (HMXBs), where massive stars waltz with compact objects.
From the dynamics of X-ray emissions to the significance of studying HMXBs in extreme environments,
this article unveils cosmic phenomena. G’s expertise in Astronomy & Astrophysics shines through,
providing a glimpse into the intriguing world of high-mass cosmic partnerships.
6. Dragonflies: Agile Flyers in the Biological Ballet Geetha Paul takes us on a mesmerizing journey into
the agile world of dragonflies, uncovering the secrets of their remarkable biology and adaptations. From
flight mechanics to predation strategies, Paul’s exploration showcases the exceptional features that have
allowed dragonflies to maintain their top predator status for millions of years.
7. The Art of Presentation: Bridging Science and Communication Closing the edition, Dr. Arun Aniyan
invites us to reflect on the art of presentation. From the three-act narrative to the balance between text
and visuals, Aniyan provides insights into effective communication, emphasizing its indispensable role
in scientific and business success.
As we celebrate the 10th edition of airis4D Journal, we invite you to immerse yourself in the richness of
these scientific narratives, each contributing a unique thread to the tapestry of knowledge. Happy reading!
iii
Contents
Editorial ii
I Artificial Intelligence and Machine Learning 1
1 Introduction to Quantum Machine Learning - Part 1 2
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 History of QML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Why Quantum Computing? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Unlocking Clustering Methods: Finding Patterns in Data 5
2.1 Partitional Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Hierarchical Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Density-based clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Model based clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Grid based Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
II Astronomy and Astrophysics 12
1 Stellar Symphony: Unveiling the Cosmic Orchestra of Spectral Classification 13
1.1 The Spectral Alphabet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 The Significance of Spectral Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Morphological Classification after Hubble 16
2.1 Morgan’s Yerkes System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 de Vaucouleurs Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Elmegreen’s Classification of Spiral Arms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 van den Bergh Classification of Galaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 High-mass X-ray binaries: Shining a Light on Cosmic Partnerships 21
3.1 High-mass X-ray binary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Importance of the study of High-mass X-ray binaries . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Some examples of High-Mass X-ray binaries . . . . . . . . . . . . . . . . . . . . . . . . . . 23
III Biosciences 26
1 The Insane Biology and Adaptations Of The Agile Flyers - The Dragonflies 27
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2 Dragonfly Flight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.3 Dragonfly Wing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
CONTENTS
1.4 The 360
O
Vision in Dragonflies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.5 Interception - a mode of hunting strategy in dragonflies . . . . . . . . . . . . . . . . . . . . . 31
IV General 33
1 The Art of Presentation 34
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.2 Orchestrate a story . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.3 Visually pleasing content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.4 A picture speaks a thousand words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.5 Voice and Eye contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
v
Part I
Artificial Intelligence and Machine Learning
Introduction to Quantum Machine Learning
- Part 1
by Blesson George
airis4D, Vol.1, No.10, 2023
www.airis4d.com
1.1 Introduction
Quantum Machine Learning (QML) is a field of research that delves into the convergence of quantum
computing and machine learning. Its primary goal is to leverage the potential of quantum computing to improve
and accelerate machine learning procedures. In a more detailed explanation, Quantum Machine Learning entails
employing quantum devices, such as quantum computers, to accelerate and enhance the functions carried out by
conventional computer programs using specialized algorithms. The inherent information processing capabilities
of quantum technologies contribute to the enhancement and expeditious execution of tasks performed by machine
learning models.
Quantum Machine Learning is currently in its nascent stages, with practical implementation remaining
limited. Nonetheless, extensive theoretical research is underway in this domain. Many professionals and
researchers are actively engaged in endeavors to craft novel machine learning algorithms and construct Quantum
Neural Networks, showcasing the fervent pursuit of advancements in the field.
1.2 History of QML
The inception of Quantum Machine Learning (QML) can be traced back to the early 2000s when D.
Ventura and Martinez [1], along with Trugenburger [2] in 2001, initially introduced foundational concepts
while conducting research on quantum memories. This early groundwork laid the theoretical foundation for
what would become a dynamic and transformative field. In 2003, E. Sch¨utzhold’s pivotal paper, titled ’Pattern
recognition on a quantum computer’,[3] made significant contributions, further solidifying the theoretical basis
of QML.
However, it was during the subsequent decade that QML experienced substantial growth and development.
This period marked a turning point with the actual creation of functional prototypes of quantum machines,
showcasing the practical potential of quantum computing for machine learning tasks. Simultaneously, the
field saw a proliferation of publications presenting a wealth of theoretical insights and experimental findings,
reinforcing the notion that QML was not merely a theoretical concept but an active and burgeoning discipline.
One notable aspect of this transformative period was the adaptation of classical machine learning algorithms
to incorporate quantum technologies. This integration was driven by the recognition that quantum computing
1.3 Why Quantum Computing?
held the promise of outperforming classical counterparts in various machine learning applications.
Within this dynamic landscape, a range of groundbreaking algorithms and techniques emerged. These
included the Quantum Support Vector Machine (QSVM) [4] [5], designed to enhance the capabilities of support
vector machines using quantum computing. Quantum Algorithms for Nearest-Neighbor Methods[6] contributed
to the advancement of pattern recognition and data analysis. Quantum Encoders [7] enabled the efficient
representation of classical data in quantum states, unlocking the potential for quantum data processing. Quantum-
Enhanced Feature Spaces expanded the capacity to capture intricate patterns within datasets. Additionally,
researchers ventured into the integration of artificial feed-forward neural networks into a quantum framework,
unveiling new possibilities for neural network applications. The era also witnessed the arrival of Deep Quantum
Neural Networks, which exemplified the vast potential of quantum computing in the training and operation of
deep neural networks.
These developments collectively marked a dynamic and transformative era in QML, characterized by active
exploration and the development of a diverse array of quantum algorithms and techniques aimed at enhancing
machine learning capabilities, ultimately reshaping the landscape of both quantum computing and machine
learning.
1.3 Why Quantum Computing?
Quantum mechanics, a branch of physics that emerged in the early 1900s, was originally developed to
elucidate the behaviors of atoms and particles on a minuscule scale. Its profound insights subsequently paved
the way for transformative technological advancements, including transistors, lasers, and magnetic resonance
imaging.
The notion of uniting quantum mechanics with information theory first surfaced in the 1970s, but it
remained relatively obscure until 1982 when physicist Richard Feynman delivered a compelling lecture. In
his discourse, Feynman contended that classical logic-based computing faced insurmountable challenges when
attempting to process calculations describing quantum phenomena efficiently. He proposed that computing
systems grounded in quantum phenomena, designed to replicate other quantum phenomena, could potentially
circumvent these limitations. Although this visionary concept laid the groundwork for what would later become
the field of quantum simulation, it did not initially ignite widespread research efforts.
Quantum and classical computers share the overarching goal of problem-solving, yet their approaches to
data manipulation and problem-solving are inherently distinct. This section unveils the unique characteristics
of quantum computers by introducing two foundational principles of quantum mechanics that underpin their
functionality: superposition and entanglement.
Superposition, a seemingly paradoxical attribute of quantum objects like electrons, enables them to occupy
multiple ”states” concurrently. For instance, an electron can exist simultaneously in the lowest energy level
of an atom and the first excited level. When an electron is prepared in such a superposition, it possesses a
probability of being in either the lower or upper state. Only through measurement does this superposition
collapse, definitively placing the electron in one of the two states.
Understanding superposition provides the basis for comprehending the fundamental unit of information in
quantum computing: the qubit. In classical computing, bits, represented by transistors, have binary states of
0 and 1, signifying off and on. In contrast, qubits, exemplified by electrons, assign 0 and 1 to states similar
to the lower and upper energy levels discussed earlier. Qubits distinguish themselves from classical bits by
their ability to exist in superpositions with varying probabilities, subject to manipulation by quantum operations
during computational processes.
3
BIBLIOGRAPHY
Entanglement, another enigmatic phenomenon within quantum mechanics, arises when quantum entities
become interconnected to the extent that none can be characterized independently of the others. Individual
identities become indistinguishable. The challenge lies in conceptualizing how entanglement can persist across
significant distances. Notably, when one component of an entangled pair undergoes measurement, it instanta-
neously determines the measurements of its partner, seemingly suggesting the transmission of information at
speeds surpassing that of light.
In recent years, the prospect of quantum computing offering faster and more cost-effective solutions
to complex combinatorial problems has spurred investments totaling billions of dollars. Quantum computing
holds the potential to revolutionize artificial intelligence, a domain that frequently entails intricate combinatorial
data processing on an immense scale to enhance predictions and decision-making, encompassing applications
like facial recognition and fraud detection. An expanding realm of research known as quantum machine
learning explores how quantum algorithms can accelerate AI processes. Nevertheless, prevailing constraints
in technology and software render the realization of quantum artificial general intelligence a relatively distant
prospect—though it undeniably transforms the concept of sentient machines from the realm of science fiction
into a tangible pursuit.
Bibliography
[1] Ventura, D., and Martinez, T. (2000). Quantum associative memory. Information sciences, 124(1-4), 273-
296.
[2] Trugenberger, C. A. (2001). Probabilistic quantum memories. Physical Review Letters, 87(6), 067901.
[3] Sch¨utzhold, R. (2003). Pattern recognition on a quantum computer. Physical Review A, 67(6), 062311.
[4] Rebentrost, P., Mohseni, M., and Lloyd, S. (2014). Quantum support vector machine for big data classifica-
tion. Physical review letters, 113(13), 130503.
[5] Willsch, D., Willsch, M., De Raedt, H., Michielsen, K. (2020). Support vector machines on the D-Wave
quantum annealer. Computer physics communications, 248, 107006.
[6] Wiebe, N., Kapoor, A., Svore, K. (2014). Quantum algorithms for nearest-neighbor methods for supervised
and unsupervised learning. arXiv preprint arXiv:1401.2142.
[7] Romero, J., Olson, J. P., and Aspuru-Guzik, A. (2017). Quantum autoencoders for efficient compression of
quantum data. Quantum Science and Technology, 2(4), 045001.
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 different domains.
4
Unlocking Clustering Methods: Finding
Patterns in Data
by Jinsu Ann Mathew
airis4D, Vol.1, No.10, 2023
www.airis4d.com
Clustering is a fundamental technique in data analysis and machine learning that involves grouping similar
data points together based on certain criteria or patterns. It’s a process of discovering natural groupings or
clusters within a dataset, where data points within the same cluster are more similar to each other than they are to
data points in other clusters (Figure 1). Clustering analysis is the systematic exploration of data to identify these
clusters. It helps researchers and analysts uncover hidden structures within their data, gain insights, and make
data-driven decisions. By organizing data into meaningful clusters, clustering analysis can simplify complex
datasets, enhance data interpretation, and facilitate further analysis or decision-making processes.
One can gain insight into this by examining the following example. Imagine you have a large dataset
containing information about customers of an e-commerce platform. This dataset includes details like age,
income, browsing behavior, and purchase history. Now, you want to understand your customer base better to
tailor marketing strategies and product recommendations. This is where clustering comes into play.
Clustering is like sorting your customers into different groups based on similarities. It’s as if you’re trying to
find distinct segments within your customer base. For instance, you might want to identify groups of customers
who tend to buy similar products, have similar spending habits, or share common interests. Clustering helps
you accomplish this by automatically grouping customers with similar attributes together, creating clusters or
segments.
Clustering analysis is the systematic process of performing this grouping or clustering. It involves using
mathematical algorithms and techniques to find patterns and similarities within the data. In our e-commerce
example, clustering analysis could help you discover that your customer base naturally divides into clusters,
such as ”Frequent Shoppers,” ”Occasional Shoppers,” and ”Budget Shoppers.” By doing so, you gain valuable
insights into your customer demographics and behaviors, which can inform marketing campaigns, product
recommendations, and inventory management.
There are several different approaches to clustering, each with its own strengths and weaknesses(Figure 2).
Some of the common clustering methods include:
Partitioning-Based Clustering:This approach partitions the data into distinct clusters.
Hierarchical Clustering: Hierarchical clustering creates a tree-like structure (dendrogram) that represents
the relationships between data points at different levels of granularity.
Density-Based Clustering: In density-based clustering, clusters are formed based on the density of data
points in the feature space.
(image courtesy: https://www.javatpoint.com/clustering-in-machine-learning )
Figure 1: Working of the clustering algorithm. Various fruits are clearly categorized into multiple clusters
sharing similar characteristics.
(image courtesy:https://link.springer.com/chapter/10.1007/978-981-13-7403-6 9)
Figure 2: Categorization of clustering algorithms
6
2.1 Partitional Clustering
(image courtesy: https://computing4all.com/courses/introductory-data-science/lessons/a-few-types-of-clustering-algorithms/)
Figure 3: Partitional clustering algorithm
Model-Based Clustering: Model-based clustering assumes that the data is generated from a statistical
model, such as a Gaussian mixture model. It aims to fit such models to the data and use them to identify clusters.
Grid-Based Clustering: Grid-based clustering divides the data space into a grid structure and assigns
data points to grid cells. Clusters are then formed based on the density of data points within these cells.
2.1 Partitional Clustering
Partitional clustering algorithms attempt to partition the dataset into non-overlapping subsets or clusters,
where each data point belongs to exactly one cluster. The following figure (Figure 3)shows three partitions
formed in a two-dimensional space representing a two-dimensional dataset. The points in each of the partitions
form a cluster. Partitional clustering is suitable for situations where we have prior knowledge of the number of
clusters we want to create or when we want to divide our data into a fixed number of non-overlapping groups. It’s
a popular choice for tasks like customer segmentation, image compression, and document categorization, among
others. One of the most well-known partitional clustering algorithms is K-Means. In K-Means, we specify
the number of clusters (K) we want to create, and the algorithm iteratively assigns data points to the nearest
cluster center (centroid) and updates the centroids until convergence. Other common partitional algorithms
are K modes, PAM, FCM etc. One limitation of partitional clustering is that it can be sensitive to the initial
placement of cluster centers, and the quality of the clustering result can vary depending on the choice of K and
the initialization method. To mitigate this, multiple runs of the algorithm with different initializations are often
employed.
2.2 Hierarchical Clustering
Hierarchical clustering is a different way to group data than partitioned clustering. The key difference is
that you don’t need to decide beforehand how many groups you want. Instead, it arranges the data into a tree-like
7
2.3 Density-based clustering
(image courtesy: https://harshsharma1091996.medium.com/hierarchical-clustering-996745fe656b)
Figure 4: Hierarchical Clustering
structure called a dendrogram to show relationships between data points at different levels of detail. There are
two types of approaches for the creation of hierarchical decomposition (Figure 4), they are:
Agglomerative Hierarchical Clustering: Agglomerative hierarchical clustering is a method that starts
with each data point as a separate cluster and gradually merges them into larger clusters. It begins by considering
every data point as its own cluster and then proceeds to iteratively combine the two closest clusters or data
points. This merging process continues until all data points belong to a single cluster. As clusters merge, a
dendrogram, which is a tree-like structure, is constructed to represent the hierarchical relationships among data
points. The height at which clusters merge in the dendrogram indicates their similarity or dissimilarity.
Divisive Hierarchical Clustering: In contrast to agglomerative clustering, divisive hierarchical clustering
starts with all data points in a single cluster and recursively divides or ”divisively” separates them into smaller
clusters. This approach aims to identify subclusters within a larger cluster by repeatedly splitting it based on
specific criteria, often to maximize the dissimilarity between the new clusters. Like agglomerative clustering,
divisive clustering also results in a dendrogram, but this dendrogram represents a top-down hierarchy, with
clusters progressively dividing into subclusters.
Hierarchical clustering is helpful in identifying clusters of varying sizes and shapes, making it suitable
for datasets with complex structures. Moreover, it is a flexible approach that accommodates various distance
metrics and linkage methods, allowing users to tailor the analysis to their specific needs. However, it’s important
to note that hierarchical clustering can be computationally intensive for large datasets, and the interpretation of
dendrograms can be subjective, requiring careful consideration of where to cut the tree to obtain meaningful
clusters.
2.3 Density-based clustering
A density-based clustering algorithm is like a tool that can group data points together, and it’s quite good
at handling data that doesn’t follow the usual pattern of having a center where most points gather(Figure 5).
Imagine your data as a collection of points on a graph. Normally, in a group of points, many of them
are close to a central point, and they gradually become sparser as you move away from that center, like people
8
2.4 Model based clustering
(image courtesy: https://en.wikipedia.org/wiki/DBSCAN)
Figure 5: Density based clustering
crowding around a popular spot. This is similar to what’s called a ”normal distribution.”
However, sometimes your data doesn’t behave this way. It might have clumps of points scattered all over
without a clear center. That’s where density-based clustering comes in handy. It can find and group these
scattered points into clusters, even when they don’t follow the usual crowd-around-a-center pattern. It’s like
finding hidden groups of friends in a big crowd, even if they’re not all standing in one place.
2.4 Model based clustering
Model-based clustering is a clustering technique that leverages probabilistic models to group data points
into clusters. Unlike some other clustering methods that rely solely on distance measures or density, In
model-based clustering, the underlying assumption is that the dataset is a mixture of multiple probability
distributions, each representing a cluster. These probability distributions can be Gaussian (normal), Poisson, or
other types depending on the nature of the data. The algorithm begins by estimating the parameters of these
probability distributions, including means, variances, and mixing proportions. These parameters represent the
characteristics of each cluster. Once the model parameters are estimated, the algorithm assigns each data point
to the cluster for which it has the highest probability of being generated. Model-based clustering is flexible and
can identify clusters of various shapes and sizes, making it suitable for a wide range of data distributions.
Model-based clustering is advantageous when dealing with complex data distributions or datasets where
clusters have different shapes and sizes. However, it may require a relatively large amount of data to estimate
the model parameters accurately. Additionally, the choice of the underlying probability distribution and model
selection criteria can impact the clustering results, and the algorithm may not perform well when data points
are not well-represented by the assumed mixture model.
2.5 Grid based Clustering
Grid-based clustering is a data analysis technique used primarily in spatial data mining and geographical in-
formation systems (GIS). It simplifies the process of identifying clusters and patterns within a multi-dimensional
attribute space. This method is particularly useful when dealing with datasets where the attributes represent
9
2.5 Grid based Clustering
(image courtesy: https://www.researchgate.net/figure/Ordinary-Grid-based-Clustering-Algorithms fig4 235623387)
Figure 6: Grid based clustering algorithm
continuous values, such as latitude and longitude coordinates, and traditional clustering methods may not be
efficient due to the high dimensionality of the data. Grid-based clustering achieves this simplification by di-
viding the attribute space into a grid of cells(Figure 6). Each cell represents a distinct region or portion of
the multi-dimensional space, and data points are assigned to these cells based on their attribute values. This
discretization process essentially transforms the continuous attribute space into discrete grid cells, making it
easier to analyze and discover clusters.
Once the data points are assigned to cells, grid-based clustering identifies clusters by examining the contents
of each cell and its neighboring cells. Clusters are formed by considering points in adjacent cells that meet
specific criteria, such as a minimum number of points within a cell and its neighboring cells. This process
allows for the discovery of clusters with relatively simple shapes and uniform densities.
One notable advantage of grid-based clustering is its computational efficiency, particularly in situations
where dealing with high-dimensional data can be computationally intensive. By reducing the dimensionality
through grid cell assignment, grid-based clustering significantly reduces the number of distance calculations
needed for clustering. It is also capable of handling large datasets efficiently.
However, grid-based clustering has its limitations. It may not perform well when clusters have irregular
shapes or varying densities, as it relies on predefined grid cells that may not conform to the underlying data
distribution. The choice of grid cell size can also impact the results, and determining the optimal cell size can
be a challenging task. Despite these limitations, grid-based clustering remains a valuable tool in spatial data
analysis and related fields, where its advantages in computational efficiency and interpretability often outweigh
its limitations.
Conclusion
In conclusion, clustering approaches play a pivotal role in data analysis and machine learning, enabling
the discovery of hidden patterns, structures, and insights within diverse datasets. Partitioning-based clustering
methods like K-Means provide simplicity and efficiency when the number of clusters is known, while hierarchical
clustering offers a hierarchical view of data relationships. Density-based clustering, excels at handling irregularly
shaped clusters and noise detection. Model-based clustering leverages probabilistic models to capture complex
data distributions, and grid-based clustering simplifies high-dimensional data analysis. Each approach has its
strengths and weaknesses, and the choice of clustering method should align with the specific characteristics and
10
2.5 Grid based Clustering
goals of the dataset. Ultimately, the art of clustering lies in selecting the most appropriate technique to unveil
the underlying structures, facilitating data-driven decision-making and insightful discoveries across various
domains.
References
Clustering in Machine Learning,Java T Point
Data Mining – Cluster Analysis , geeksforgeeks
Clustering — Introduction, Different Methods, and Applications ,Sauravkaushik8 Kaushik, analytics
vidhya, July,2023
A short review on different clustering techniques and their applications,Emerging Technology in Mod-
elling and Graphics: Proceedings of IEM Graph,2020
INTRODUCTION TO MACHINE LEARNING/DATA SCIENCE, Computing for all
Hierarchical Clustering,Himanshu Sharma, medium,April, 2021
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.
11
Part II
Astronomy and Astrophysics
Stellar Symphony: Unveiling the Cosmic
Orchestra of Spectral Classification
by Robin Jacob Roy
airis4D, Vol.1, No.10, 2023
www.airis4d.com
The universe is an astronomical marvel, teeming with countless celestial objects and cosmic phenomena,
with stars taking center stage as the most prominent players in the celestial theater. These celestial luminaries
shine brightly across the vast expanse of space, displaying a remarkable range of sizes, temperatures, and
compositions. To fathom the diversity of stars and to unravel the complexities of their nature, astronomers rely
on a system known as spectral classification.
Stellar classification is a structured method for categorizing stars based primarily on their temperatures,
which are primarily inferred from their spectra. This widely embraced classification approach is an amalgamation
of two primary systems: the Harvard system, which categorizes stars by their surface temperatures, and the MK
system, which hinges on their luminosity.
In the 1860s, Italian astronomer Angelo Secchi initially identified four primary spectral types of stars.
Later, at the Harvard College Observatory in the 1880s, while compiling the Henry Draper Catalogue of stars,
additional spectral types were established. They were designated with letters in alphabetical order, reflecting
the strength of their hydrogen spectral lines. Much of this classification work was carried out by three dedicated
assistants: Williamina P. Fleming, Antonia C. Maury, and Annie Jump Cannon.
As this work advanced, the spectral types were reorganized into a sequence that followed the stars’ surface
temperatures, moving from hotter to cooler. The order of stellar types from hottest to coolest is as follows: O,
B, A, F, G, K, M. To remember this sequence, a traditional mnemonic is ”Oh Be A Fine Girl [or Guy], Kiss
Me.” In some cases, additional letters have been introduced to categorize novas and rarer types of stars. To
further specify within a type, numbers from 0 to 9 are used, with higher numbers indicating cooler stars. The
hotter stars are often referred to as ”early,” while the cooler ones are termed ”late.” Figure 1 depicts the relative
size and color of Main Sequence stars arranged from O to M.
With the discovery of brown dwarfs, celestial objects that form like stars but do not undergo thermonuclear
fusion, the stellar classification system has been extended to include spectral types L, T, and Y. This expanded
classification allows astronomers to comprehensively categorize stars and other celestial objects, enhancing our
understanding of the cosmos.
1.1 The Spectral Alphabet
Figure 1: Main-sequence stars arranged (from right to left) O to M Harvard classes. Source: Wikimedia
Commons
1.1 The Spectral Alphabet
The spectral classification system arranges stars into seven primary spectral classes, each represented by
a letter from O to M. This classification system is a powerful tool that allows astronomers to distinguish stars
based on their unique characteristics.
1. Class O: The Hottest Stars - Stars of class O are the hottest in the stellar hierarchy, with surface temperatures
exceeding 30,000 Kelvin. They appear bluish-white in color and emit intense ultraviolet radiation. O-type
stars are relatively rare, massive, and short-lived, often leading to spectacular supernova explosions.
2. Class B: Bright and Blue - Class B stars are slightly cooler than O-type stars, with surface temperatures
between 10,000 and 30,000 Kelvin. They have a bluish-white appearance and are known for their
brightness. Many prominent stars, including Rigel and Spica, belong to this class.
3. Class A: Pure White Stars - Class A stars are characterized by their white appearance, with surface
temperatures ranging from 7,500 to 10,000 Kelvin. They often serve as stable and long-lasting sources
of light.
4. Class F: Yellow-White Giants - Stars of class F have surface temperatures between 6,000 and 7,500
Kelvin. They are typically larger than the Sun and exhibit a yellow-white hue.
5. Class G: The Solar Standard - Our Sun belongs to class G, with a surface temperature of around 5,500
Kelvin. G-type stars are often referred to as ”yellow dwarfs” and are considered ideal candidates for
hosting habitable planets.
6. Class K: Orange Giants and Supergiants - Class K stars are cooler than the Sun, with surface temperatures
between 3,500 and 5,000 Kelvin. They have a distinct orange or red appearance and can be much larger
than the Sun in their giant or supergiant phase.
7. Class M: Cool Red Stars - Class M stars are the coolest, with surface temperatures below 3,500 Kelvin.
They are the most common type of star in the universe and exhibit a deep red color. Many red dwarf stars,
including Proxima Centauri, fall into this category. A summary of the Harvard spectral classification
system is given in Table 1.1.
1.2 The Significance of Spectral Classification
The classification of stars based on their spectra provides valuable insights into their physical properties,
evolutionary stages, and even the potential for hosting planets and life. Some of the key information revealed
by spectral classification includes:
1. Temperature: Spectral classification is a direct indicator of a star’s surface temperature. Hotter stars are
14
1.2 The Significance of Spectral Classification
Spectral type Temperature (K) Dominant spectral features
O >30,000 Helium absorption lines
B 10,000-30,000 Helium and hydrogen absorption lines
A 7,500-10,000 Hydrogen absorption lines
F 6,000-7,500 Calcium absorption lines
G 5,200-6,000 Calcium and magnesium absorption lines
K 3,700-5,200 Potassium and sodium absorption lines
M 2,000-3,700 Titanium oxide absorption bands
Table 1.1: Summary of the Harvard spectral classification system.
associated with higher-energy blue and ultraviolet light, while cooler stars emit predominantly red and
infrared radiation.
2. Luminosity: A star’s spectral type can provide a rough estimate of its luminosity. For example, O-type
stars are exceptionally bright, while M-type stars are relatively dim.
3. Age and Evolution: The spectral class of a star offers clues about its evolutionary stage. For instance,
massive O-type stars are typically young, while red M-type stars are often older.
4. Chemical Composition: The presence of certain spectral lines within a star’s spectrum can reveal infor-
mation about its chemical composition and the abundance of specific elements.
5. Habitability: The classification of stars plays a vital role in the search for potentially habitable exoplanets.
Stars in the G-type class, like our Sun, are often prime candidates for hosting habitable planets.
Spectral classification is the astronomical Rosetta Stone that unlocks the mysteries of the cosmos, allowing
astronomers to decipher the nature of stars based on the light they emit. This classification system has not only
deepened our understanding of the universe but also guided the search for exoplanets, informed our knowledge
of stellar evolution, and unveiled the intrinsic beauty of stars in all their diversity. As we continue to explore
the heavens, the spectral classification of stars remains an essential tool in our quest to comprehend the vast and
intricate tapestry of the universe.
References:
Harvard Spectral Classification
What color are stars
Spectral Classification
Classification of stellar spectra
About the Author
Robin is a researcher in Physics, specializing in the applications of Machine Learning for Astronomy
and Remote Sensing. He is particularly interested in using Computer Vision to address challenges in the fields of
Biodiversity, Protein studies, and Astronomy. He is presently engaged in utilizing machine learning techniques
for the identification of star-forming knots.
15
Morphological Classification after Hubble
by Sheelu Abraham
airis4D, Vol.1, No.10, 2023
www.airis4d.com
The famous tuning fork diagram for galaxy classification by Hubble has gained more-or-less universal
acceptance due to its simplicity in the scheme. As the fundamental units of the Universe, it is essential
to understand the formation and evolution of galaxies. The simplest way to study the nature of galaxies is to
classify them according to their appearance, which Hubble did with his Classification Scheme. The classification
scheme provides a simple way to understand the galaxies’ nature when they were poorly understood. It also
gives a framework for the logical approach for further studies about galaxies. While the Hubble classification
appears quite simple, the high significance of his scheme emerged from the correlation between the measured
physical properties of galaxies and their positions in the sequence.
2.1 Morgan’s Yerkes System
Later, when the observed number of galaxies increases, many do not fit comfortably into the Hubble
scheme, especially at the core of the clusters. In 1957, Morgan and Mayall devised a new scheme for classifying
galaxies according to their spectroscopic characteristics [2]. This scheme in effect is based on the degree of
central concentration luminosity, one of the criteria used by Hubble. The simplified scheme of Yerks (Morgan)
classification is illustrated in Figure 1. The figure shows that the (S) and (B) sequences resemble those of the
Hubble system. But there are considerable differences between the two systems because of a lack of correlation
between the appearance of the arm structure and the nuclear region. The scheme divided the galaxies into three
Figure 1: According to the degree of central concentration of luminosity, the galaxies are classified into seven
groups in the simplified Yerkes (or Morgan) system.
2.2 de Vaucouleurs Classification
Figure 2: G
´
erard de Vacucouleurs
general areas. The first region consists of highly irregular galaxies lacking symmetrical nuclear concentration
of light. The second region includes galaxies of the most substantial central concentration of luminosity. This
class consists of giant ellipticals, elliptical-like objects which differ from pure ellipticals and barred systems.
the spectroscopic evidence indicates that these are the most highly composite systems of all. In Hubble’s
classification system, a conflict occasionally arises between the arm morphology and the central concentration
of light was rectified in the Yerkes system. Since the Yerkes system was based on a single parameter, it could
be effectively used for the classification of galaxies in very rich clusters.
2.2 de Vaucouleurs Classification
G
´
erard de Vacouleus devised another way of classification of disk galaxies. It is an extension of the
Hubble classification of galaxy types involving a ”three-dimensional” classification grid from ellipticals to
irregulars [2]. The de Vaucouleurs’ classifications are almost similar to Hubble’s classifications in the case
of elliptical galaxies. However, for other galaxies, differences are made between normal and ringed, barred
and unbarred, and transitional types. Thus, de Vaucoulers’ scheme provides a more detailed description, even
the sub-classifications based on more simple structures [1]. The scheme provides a higher resolution at the
late end of Hubble Classification. This classification system has the sequence E–S0–Sa–Sb–Sc–Sd–Sm–Im,
where the index m refers to Magellanic, which means the galaxies resemble the Magellanic clouds [1]. The
three-dimensional classification scheme is illustrated in Figure 5. Thus, his approach gave greater emphasis and
more accurate recognition of specific details like rings, bars, spiral structures etc.
2.3 Elmegreen’s Classification of Spiral Arms
Elmegreen’s classification is based primarily on visual observations of galaxies, and it attempts to capture
the diversity of spiral arm patterns that astronomers have observed. This scheme was proposed by the astronomer
Bruce Elmegreen to categorize the different morphological types of spiral arms found in galaxies. Early studies
from 1925 onwards show that some galaxies had ‘massive’ arms, whereas others exhibited a ‘filamentous’
17
2.3 Elmegreen’s Classification of Spiral Arms
Figure 3: A simplified Hubble-de Vaucouleurs classification scheme for elliptical and lenticular galaxies
Figure 4: A simplified Hubble-de Vaucouleurs classification scheme for for spiral galaxies
18
2.4 van den Bergh Classification of Galaxies
Figure 5: Diagram showing the three-dimensional version of the basic de Vaucouleurs classification scheme
Here the type A and B (unbarred and barred) classification does not apply to strictly elliptical galaxies. Also,
in my catalog, de Vaucouleurs’ type E+ is represented by E/S0, E/SA0, E/SB0, etc, as appropriate
Figure 6: Figure shows some example galaxies with spiral arm classes of Elmegreen’s classification. The
galaxies are (left to right): Row 1 - NGC 45, 7793, 5055, 2403, and 1084. Row 2: NGC 6300, 2442, 3504,
5364, and 1365.
spiral structure. More recently, Elmegreen & Elmegreen (1982, 1987) introduced a comprehensive twelve-stage
classification system for spiral arms. This system ranges from Type 1 arms, characterized by their ragged,
patchy, or disorderly nature, to Type 12 arms, distinguished by their elongated, symmetrical, and crisply defined
appearance, often dominating the galaxy’s overall look [2]. Figure 6 shows some example galaxies with spiral
arm classes of Elmegreen’s classification. It’s important to note that this classification is just one of several
attempts to describe the complex structures of spiral galaxies and how their arms are organized.
2.4 van den Bergh Classification of Galaxies
The van den Bergh galaxy classification system considered the luminosity effects of galaxies developed
by Sidney van den Bergh in 1998 [4]. It provides a means to describe and categorize the diverse population
of dwarf galaxies, which may not fit neatly into the traditional Hubble sequence designed for larger, more
prominent galaxies. Luminosity effects are evident in the morphology of galaxies through surface brightness
19
2.5 Conclusion
Figure 7: The DDO system for galaxy classification. Here the left side shows the elliptical galaxies, the upper
arm of the tuning fork represents the lenticulars, lower the spirals and the middle arm shows anemic spirals.
differences between giants and dwarfs. van den Bergh describes his classification system based on luminosity
effects using a set of luminosity classes analogous to those used for stars. Based on inspection of the prints of the
Palomar Sky Survey, van den Bergh (1959, 1966) could identify and catalogued 243 David Dunlap Observatory
(DDO) dwarf galaxies. Figure 7 shows the DDO classification of galaxies. This system of classification is
little complicated. In this class, a sequence of ’anemic spirals’ occur most frequently in rich clusters, and their
characteristics are intermediate between those of normal spirals and gas-poor systems of type S0. The variations
observed within each class are tentatively explained by considering how the surrounding environment influences
the evolutionary path of flattened galaxies.
2.5 Conclusion
The different classification schemes provide different ways to categorise the galaxies based on their
morphologies. Even though all these schemes have their way of explaining the classes, Hubble’s classification
stays as the simplest way of classification of galaxies. All the other schemes are based on Hubble’s scheme and
tried to be revised as the observation capabilities are enhanced.
Bibliography
[1] Ronald J. Buta, Harold G. Corwin, and Stephen C. Odewahn. The de Vaucouleurs Atlas of Galaxies. 2007.
[2] Sidney Bergh. Galaxy Morphology and Classification. Cambridge University Press, 1998.
[3] Ronald J. Buta. Galaxy morphology, 2011.
[4] S. van den Bergh. A new classification system for galaxies. , 206:883–887, June 1976.
About the Author
Dr Sheelu Abraham is an Assistant Professor at Department of Physics, Mar Thoma College,
Chungathara and a Visiting Associate at Inter University Center for Astronomy & Astrophysics, Pune. Her area
of specialisation is ML applications for Astronomy.
20
High-mass X-ray binaries: Shining a Light on
Cosmic Partnerships
by Sindhu G
airis4D, Vol.1, No.10, 2023
www.airis4d.com
3.1 High-mass X-ray binary
High-Mass X-ray Binaries (Fig: 1) are a class of binary star systems that consist of a massive star and a
compact object. This compact object has formed from a massive star that underwent a supernova explosion at the
end of its life cycle. The remnants of the explosion can become a highly dense object with strong gravitational
pull. This compact object can be either a neutron star or a black hole. In these systems, the massive star is
typically an O or B-type star, which means it is much more massive and hotter than the Sun. In a high-mass
X-ray binary, the companion star has a mass greater than 10 solar masses. The term ”X-ray” in their name refers
to the fact that they emit X-rays, which makes them detectable by X-ray telescopes.
The X-ray emissions in high-mass X-ray binaries originate from the process of material transfer from a
massive star to a neutron star or black hole. The massive star loses its mass via a potent stellar wind, and a
portion of this material is captured by the compact object. As the material spirals in towards the compact object,
it forms a hot accretion disk. As this material descends onto the compact object, it undergoes intense heating,
leading to the emission of X-rays.
While high-mass X-ray binaries are known for their significant X-ray emissions, it is worth noting that
a considerable portion of them can be observed in the visible light spectrum. This is because their optical
emissions are primarily influenced by the presence of the massive companion star.
HMXBs often exhibit variability in their X-ray emissions. This variability can arise from changes in the
accretion rate, variations in the geometry of the accretion disk, or other factors. The study of X-ray variability
can provide insights into the dynamics of these binary systems. The changes in X-ray luminosity is connected
to variations in the strength and fluctuations in the velocity of the stellar wind, as well as changes in the
separation distance between the binary system’s component stars. Variability within HMXBs is evident through
phenomena such as X-ray pulsars rather than X-ray bursters. Notably, HMXBs generally display a greater X-ray
luminosity in comparison to LMXBs.
High-mass X-ray binaries (HMXBs) are found in various regions of our Milky Way galaxy, as well as in
other galaxies in the universe. Approximately 152 HMXBs have been discovered in the Milky Way. HMXBs
are typically found in regions of active star formation, such as stellar nurseries and spiral arms of galaxies,
where massive stars are more common. Studying these systems helps astronomers understand the life cycles of
massive stars, the physics of accretion, and the impact of high-energy processes on their surroundings.
3.2 Importance of the study of High-mass X-ray binaries
Figure 1: An artist’s impression of a High Mass X-ray Binary, Cygnus X-1 Source: NASA/CXC/M.Weiss.
HMXBs, despite their relatively short lifespans of just a few tens of millions of years, are not expected to
have strayed far from their initial birthplaces. The reason for the separation between these systems and their
birth sites can be attributed largely to the momentum imparted during the explosive supernova event of the more
massive star in the binary pair. The identification of these stars’ birthplaces plays a pivotal role in evaluating
the influence of the surrounding environment on the formation and subsequent evolution of these systems.
3.2 Importance of the study of High-mass X-ray binaries
HMXBs are of great interest to astronomers and astrophysicists because they offer a unique opportunity
to study extreme physical conditions, including strong gravitational fields and high temperatures. They also
provide insights into the evolution of massive stars, the formation of compact objects, and the behaviour of
matter in these extreme environments. The interaction between the massive star and the compact object can lead
to mass transfer and the eventual evolution of the massive star, possibly resulting in a supernova explosion. If
the compact object is a neutron star or a black hole, the study of HMXBs can help probe the extreme conditions
of strong gravity and dense matter. HMXBs are bright sources of X-rays, making them essential for X-ray
astronomy. Observing HMXBs allows astronomers to study X-ray emission processes, the behaviour of matter
under extreme conditions, and the properties of compact objects. Some HMXBs are associated with gamma-ray
bursts, which are highly energetic and short-lived events thought to be associated with the formation of black
holes. These bursts are among the most powerful explosions in the universe. HMXBs can release significant
amounts of energy and matter into their surroundings through their intense radiation and powerful stellar winds.
This can influence the structure and evolution of their host galaxies.
3.3 Classification
Based on variations in their observational characteristics, particularly the type of companion star, HMXBs
can be categorised into two main groups: supergiant X-ray binaries (SGXBs) featuring an OB supergiant as the
secondary star and Be X-ray binaries with a Be star as the primary component. Supergiant X-ray binaries, the
former category, make up roughly one-third of all the HMXBs identified within the Milky Way. In contrast, Be
X-ray binaries, the latter category, constitute approximately three-fifths of these binary systems. In BeHMXBs,
22
3.4 Some examples of High-Mass X-ray binaries
Figure 2: Chandra X-ray Image of Cygnus X-1 Source: NASA/CXC.
the mechanism of material transfer involves the interaction between a compact object and a decretion disk.
On the other hand, sgHMXBs primarily transfer mass through the presence of a vigorous stellar wind. In
exceptional cases, accretion in sgHMXBs might occur via Roche-lobe overflow, resulting in higher X-ray
luminosities compared to systems that accrete via stellar winds.
3.4 Some examples of High-Mass X-ray binaries
Here are a few examples of high-mass X-ray binaries (HMXBs):
3.4.1 Cygnus X-1
Cygnus X-1 (Fig: 1, Fig: 2, and Fig: 3) is one of the most famous HMXBs and was the first strong black
hole candidate ever discovered. It consists of a massive blue supergiant star and a black hole. This binary
system emits copious amounts of X-rays and is located in the constellation Cygnus.
3.4.2 Vela X-1
Vela X-1 (Fig:4) is an HMXB that contains a neutron star. It consists of a massive B-type supergiant star
and a neutron star. This system emits intense X-ray radiation and is located in the Vela constellation.
3.4.3 Cyg X-3
Cygnus X-3 (Fig:5) is an HMXB with an exotic, highly compact object. It is believed to contain a neutron
star or a black hole. Cygnus X-3 is known for its rapid X-ray and radio variability and is an active source for
multi-wavelength observations.
23
3.4 Some examples of High-Mass X-ray binaries
Figure 5: Cygnus X3 and its X-ray halo Source: NASA.
References:
High-mass X-ray Binaries
High-Mass X-ray binary: Classification, Formation,and Evolution
Variability in High Mass X-ray Binaries
WATCHDOG: A COMPREHENSIVE ALL-SKY DATABASE OF GALACTIC BLACK HOLE X-RAY
BINARIES
X Ray Binaries Monitoring
X-ray Binary
Black Holes and X-ray binaries
A catalogue of high-mass X-ray binaries in the Galaxy: from the INTEGRAL to the Gaia era
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.
25
Part III
Biosciences
The Insane Biology and Adaptations Of The
Agile Flyers - The Dragonflies
by Geetha Paul
airis4D, Vol.1, No.10, 2023
www.airis4d.com
1.1 Introduction
Dragonflies are ancient insects with a remarkable evolutionary history and have developed a set of incredible
adaptations that define their biology. This article explores various facets of dragonfly biology, focusing on their
flight adaptations, wing structure, predation techniques, etc., showcases their resilience and adaptability across
various environments. Dragonflies are celebrated for their aerial prowess, including synchronised wing stroking,
enabling them to rapidly hunt their prey. Their unique hunting strategies, primarily interception, are discussed,
shedding light on their efficient predation of smaller prey. Their preference for small prey is explored within
an ecological context. Very few insects are more successful than the dragonflies, who have remained relatively
unchanged for hundreds of millions of years. With the killer state of adaptation, it helps to maintain a top
predator status. It’s no wonder, engineers and scientists are keen to study and emulate them. Dragonflies
have inspired many biomimicry projects that aim to replicate how they track their targets and perform aerial
acrobatics. Today there are no free flying terms of dimensions comparable to the dragonfly, the agile flyer. This
comprehensive exploration of dragonfly biology highlights these fascinating insects’ intricate and awe-inspiring
aspects.
1.2 Dragonfly Flight
Dragonflies exhibit a striking flight adaptation that distinguishes them from many other insects: they
possess direct-flight muscles, a feature distinct from the indirect-flight muscles prevalent in most insect
species. In typical adult insects, flight muscles are directly anchored to the thorax wall, causing the thorax to
contract and expand. This expansion induces resonance in the wings, leading to their vibration and initiating the
flapping motion. In contrast, dragonflies have flight muscles directly attached to the bases of their four wings
independently in three axes, meaning they can move in horizontal, vertical and original motion. Their four
wings can flap unparalleled, and they can propel themselves like a helicopter in all six directions- Up, down,
left, right, forward and even backwards.
Dragonflies can flap their fore and hind wings in different phases. This is called counter-stroking, where
they can stroke their fore and hind wing 180 degrees out of phase with each other. This can engage them to
have a very slow pace forward flight. In phased stroking, the hind wings flap 90 degrees ahead in phase of the
1.3 Dragonfly Wing
Figure 1: Image-courtesy: https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-8#fpstate=ive&vld=cid:
45f5140c,vid:iJi61NAIsjs,st:0
Figure 2: Figure showing the Pterostigma on wing
forewings. This helps in the fast-forward flight, far or less lift, generating a lot more thrust. This enables some
species of dragonflies to fly fifty kilometres per hour, making them fly as the fastest flying insects. Dragonflies
can also engage in synchronised stroking, where the forewings and hindwings flap together simultaneously.
This is more useful in increasing the acceleration of the flight and also in preparing for changes in directions
quickly. They use this configuration while chasing the prey. Dragonflies are also capable of flying backwards,
the backward flight. During this flight, they tilt their body up to 90 degrees upwards like some species of
hummingbirds do. Dragonflies can also glide free with the wind without flapping their fore and hind wings.
Also, some female dragonflies glide free and make a ride, getting attached to male dragonflies while mating.
Switching between such a wide variety of flight behaviours and aerial acrobatics makes dragonflies an efficient
predator of hunting flying insects and feeding predators of their own.
1.3 Dragonfly Wing
Dragonfly wings also have a prominent pterostigma, which is a small pigmented part at the leading edge
of the wing; it is heavier than the rest of the wing and acts as a counterweight.
From a biomechanics perspective, the weight of the pterostigma displaces the chordwise centre of mass
ahead the torsion axis at the level of pterostigma.
The outer edge shape of the dragonfly wing is shaped to perform the flapping movement and feathery
28
1.3 Dragonfly Wing
Figure 3: Figure showing the torsion axis
Image-courtesy: https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-8#fpstate=ive&vld=cid:45f5140c, vid:
iJi61NAIsjs,st:0
Figure 4: Pterostigma of the wing acts as a Counterbalance
Image-courtesy: https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-8#fpstate=ive&vld=cid:45f5140c, vid:
iJi61NAIsjs,st:0
29
1.4 The 360
O
Vision in Dragonflies
Figure 5: Marked layout of the veins in the wing of dragonfly.
Image-courtesy: https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-8#fpstate=ive&vld=cid:45f5140c, vid:
iJi61NAIsjs,st:0
vibrations called fluttering. Such aeroelastic flutter can be performed by dragonflies because of the small spot
on the wings which comes about 0.1% of its total body weight. This allows them to safely reach about 10 - 20%
higher flight speed.
Wings of Dragonflies are not mere flap plates; instead, they have veins with three dimensional corrugations
that make the wings performance more efficient from a structural perspective. These wing vein patterns maintain
their body- balance and prevent them without whooping or falling down.
1.4 The 360
O
Vision in Dragonflies
It’s not the speed, incredible ability of flight acrobatics alone that make dragonflies the greatest hunters in
the air, it’s how they see the world that makes all the agility possible. For that they have characteristic compound
eyes, with 30,000 individual facets or ommatidia, which are like mini telescopes which can detect light from
the direction in which they are pointed.
Figure 6: Shows the compound eyes of dragonflies and its 360
O
field of vision
Image-courtesy:https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-8#fpstate=ive&vld=cid:45f5140c, vid:
iJi61NAIsjs,st:0
As their eyes wrap around their head almost entirely, the field of vision is nearly 360
O
and it is perfect for
surveying its targets of its surroundings, while they are at perch.
30
1.5 Interception - a mode of hunting strategy in dragonflies
Figure 7: Shows the ommatidia of the compound eyes of dragonfly.
Image-courtesy:
https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-8#fpstate=ive&vld=cid:45f5140c, vid:iJi61NAIsjs,st:0
Each ommatidia has five to seven opsins. Opsins are light sensitive proteins, which can absorb different
colours of light in the spectrum and allows dragonflies to perceive colour thereby useful for hunting their prey.
1.5 Interception - a mode of hunting strategy in dragonflies
In dragonflies, interception behaviour refers to their ability to navigate and fly directly towards a point
where they can intercept and capture their prey. This behaviour involves precise control of their flight path to
minimise the relative motion of the prey’s image on their retinas. This interception capability is likely guided
by specific neural mechanisms and visual responses that enable dragonflies to successfully capture their prey
during flight.It has been observed that dragonflies exhibit a flight pattern where they navigate directly towards
the point of prey interception by skillfully adjusting their flight to minimise any displacement in the image
of the prey on their retina. The interception strategy described here for dragonflies bears a resemblance to
the approach employed by humans, particularly baseball players tracking a flying ball. It appears that mature
dragonflies employ a tactic aimed at constraining retinal slip along an upwardly extending linear path (McBeath
et al. 1995). This particular strategy results in a curved trajectory leading to the point of capture. It’s worth
noting that outfielders are confined to the two-dimensional plane of the baseball field, whereas flying dragonflies
navigate through the full three dimensions.
The dragonfly employs a series of precise visual and motor processes to capture prey. Initially, it detects
prey within a ∼90
O
cone of space and rapidly orients its gaze with a 50 ms head movement. Over the next 250
ms, smooth pursuit head movements further refine this focus until prey are within ±4
O
of the direction-of-gaze.
Once certain criteria are met, such as acceptable foveation error, overhead crossing, and prey size and speed, the
dragonfly initiates takeoff, timed to intercept the prey directly overhead. After takeoff, it uses head movements
to maintain focus on the prey and align its body and flight direction with the prey’s path. Flight termination can
occur at any point before capture.
Consequently, dragonflies possess the ability to adjust their flight to minimise retinal slip in all directions.
When pursuing prey following a linear course, this strategy culminates in a direct approach. This closed-
loop strategy implies that, for both humans and dragonflies, the pursuer does not set a course directly for a
predetermined interception point but instead synchronises their path with the flying object’s arrival.
31
1.5 Interception - a mode of hunting strategy in dragonflies
Figure 8: Schematic representation of the ‘ interception’ strategy of prey capture in dragonflies.
Image-courtesy: https://www.sciencedirect.com/science/article/pii/S0960982217302798#:
∼
:text=The%20dragonfly%20intercepts%20flying%20prey,acuity%20fovea%20on%20the%20eye
In humans, capturing a flying object is unmistakably a learned skill, while in dragonflies, the interception
strategy is more likely to be innate, deeply ingrained in their nervous system. Upon emerging from their aquatic
nymphal habitat, adult dragonflies demonstrate the capacity not only to take flight but also to orient themselves
towards objects in their surroundings, such as a convenient perch.
All the above adaptations allow dragonflies to perform intricate aerial manoeuvres, making them formidable
hunters and flyers. The dragonfly’s exceptional biology, from its flight mechanics to its predation strategies,
continues to inspire wonder and admiration among scientists and nature enthusiasts alike.
References
Baird JM, May ML (1997) Foraging behaviour of Pachydiplax longipennis (Odonata: Libellulidae). J
Insect Behav 10: 655±678 https://link.springer.com/article/10.1007/BF02765385
McBeath MK, Shaer DM, Kaiser MK (1995) How baseball outfielders determine where to run to catch fly
balls. Science 268: 569 - 573
https://www.researchgate.net/publication/12609048 Prey pursuit and interception in dragonflies
https://www.google.com/search?q=world+od+dragonflies&oq=&aqs=chrome.0.69i59i450l8.14461909j0j1&sourceid=chrome&ie=UTF-
8#fpstate=ive&vld=cid:45f5140c,vid:iJi61NAIsjs,st:0
https://www.sciencedirect.com/science/article/pii/S0960982217302798#:
∼
:text=The%20dragonfly%20intercepts%
20flying%20prey,acuity%20fovea%20on%20the%20eye
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.
32
Part IV
General
The Art of Presentation
by Arun Aniyan
airis4D, Vol.1, No.10, 2023
www.airis4d.com
1.1 Introduction
Knowledge sharing is a vital part of academia and industry. One can have groundbreaking and innovative
ideas, but if not able to share and present it is of no good. The sharing of ideas and knowledge is what makes
technology accessible to the public and thereby be more innovative.
In an academic setting, research is disseminated through presentations and journals. Scientific conferences
are more vibrant and responsive when not just some complicated scientific ideas are presented but when
presented compellingly, making it easy for people to understand. People tend to ask more questions when the
proposed idea is comprehensible and makes them think.
Similarly in an industrial setting having a brand new product with features or new business ideas is not
enough. One should be able to present that product in a convincing and attractive manner. Business presentations
have to be very compelling and simple for people to understand. In many cases, it is hard to sell any product
without a compelling presentation.
Presentation skills are now a basic requirement for many job requirements and organizations tend to pick
people who are good at presenting ideas. This is not just a professional skill but also a very important social
skill to connect with people around you.
In industry when a company launches a new product or even features to their existing product they do
it through a highly publicised event. Apple events are the best example of this and they started the trend of
such events. They would get their CEO and their best speakers to launch and explain their new product. The
presentations they give are very engaging and watched by large masses. Figure 1 shows the event when Steve
Jobs introduced the iPhone. The tag he uses is “Apple re-invents the phone” which is the starting slide. This
model has now been followed by most large organizations. The impact that these presentations make on the
market and the public space is huge. It is to be noted this works irrespective of the brand value of organizations.
1.2 Orchestrate a story
Storytelling is the simplest and oldest form of presentation. Everyone across all age groups likes to hear
stories. Stories connect people and their brains to the storyteller. It is a powerful tool to synchronize human
brain activity.
Every story has three basic components - (1) Setup, (2) Conflict, and (3) Resolution ??. This is commonly
referred to as the three-act narrative of a story. Basically what this means is that there is a beginning premise
1.2 Orchestrate a story
Figure 1: Steve Jobs presenting the Apple iPhone for the first time. He is considered the best public speaker
in the public space. His method of presentation has been later copied by many organizations for their product
launches.[Image credit: Google]
which is very normal for all stories. In the second stage, something goes wrong out of the blue. This is where
the story narrative starts to peak and people’s minds are unconsciously eager to know what happens next. This
part glues people to their seats to know how the conflict gets resolved. The resolution finally solves the conflict
which is the most interesting part of the story and gives the listeners a sense of excitement and happiness. This
is when you usually have a happy ending.
Figure 2: Illustration of the 3-point narrative where the curve shows the level of engagement/excitement of the
story.[Image credit ]
Every story has a roller-coaster effect, where it starts gradually and slowly peaks up the excitement and
35
1.3 Visually pleasing content
suddenly drops down to fill in the adrenaline and finally stops at ease. Presentations can also take this model
and make the experience engaging.
1.2.1 Setup
This simple rule can be applied to presentations as well. The first question one needs to ask is why I am
telling this story and why should my audience care about it. This is where you set the premise of the presentation.
There must be a compelling reason why a piece of research was conducted or a product was developed. That
same state of excitement should be shared and set when you start the presentation. The key to this is to set
up a premise and try to bring your audience to the same state of mind that you were in at that moment. Let
the audience see and feel the world the same way you did. It is important to use the right words and tone to
bring your audience to see what you want them to. Once this is properly set, you have the audience with you.
Emphasizing specific terms and words that reflect the idea of your topic during the setup phase is important to
keep the continuity. This will help to further resonate the key points of the topic throughout the presentation.
One method is to start the presentation with a completely off-topic but related slide in the beginning. It may
be a picture or even a thought that quickly catches the attention of your audience and can be brought back as a
closing note at the end.
1.2.2 Conflict
The conflict is often the crux of your presentation that makes your audience stick to the narrative. Unex-
pected conflicts often make the story more interesting. In the case of a presentation, once the background is set,
the challenging part of the problem needs to be presented. This is similar to highlighting the deficiencies of
existing methods and technologies in a scientific paper’s introduction. Convincing your audience that conflict
is necessary for the story to progress to the next level. In this phase, the presentation should focus on why the
new idea you are bringing forward is worthy of attention. Make sure continuity is met with the ‘setup’ phase
and progression is natural. The audience should never feel that you jumped onto to completely different topic.
The conflict phase should be the preamble to your selling point. In a business environment, people should be
already convinced that should they be paying you their money. For an academic, you might be pointing out why
we need a new solution for an existing problem.
1.2.3 Resolution
This is the part where you should be able to take your audience’s excitement to the peak. Now that the
setup and conflict are clear, the resolution should be the happy ending that everyone wants. For academics, one
basically shows the results and findings of the research conducted. It is important to show only those figures and
numbers that your audience desires and nothing more to avoid confusion. Show them exactly what you want
them to see and focus on. As a common anecdote “only show the pleasing results”. In a business presentation,
one would show how the new product would save people money and time or even make their lives better than
before.
Structuring presentations in a three-act narrative will automatically help orchestrate the different parts of
the idea into a story and make it very engaging for people to listen to you. The key point here is to drive your
audience’s thoughts through your lens so that they have the same emotions as you would. This will make your
presentation more engaging and connected with the audience.
36
1.3 Visually pleasing content
1.3 Visually pleasing content
Presentation material which is commonly put on slide decks, is the next key component to making an
engaging presentation. The number of slides also is the first factor that should be considered when orchestrating
a presentation. One should not flood the audience with hundreds of slides. The ideal rule is to keep the number
of slides to half the number of minutes of the presentation. For example, for a 40-minute talk, there should be
no more than 20 slides at maximum including the title slide. Flashing too much information over a time period
will not help register the message that you sharing.
The colours and shades of the slides impact the tone of your presentation. Colours directly translate to
emotions in the brain and therefore the colour theme for the presentation should be properly chosen. The simple
rule is to avoid very bright coloured backgrounds and choose light ones. Often when reading printed material,
text on a white background is easier for the eyes to read. But when on a screen, it is easier for people to read
white text on a dark background. The “dark mode” colour theme on phones and computers is geared towards
this idea.
On a similar note, try to avoid box-type shapes in the slides. They are more old-fashioned as well as not
elegant in terms of display. Figure 3 shows an example with a box-type graphic. This is visually simplistic but
the brain does register it as aesthetically pleasing.
Figure 3: A sample slide with a box-type design and very bright colour. This is an example where the slide
does not conform to a visually pleasing style.
On the other hand, figure 4 shows a different approach to the previous slide with more subtle colours and
without the box-type design.
When it comes to presenting slides through a projected screen, environmental lighting is a major factor. If
it is possible to know the lighting setup of the presentation area beforehand, then choose a theme that fits the
lighting there. If the lighting condition is one in which there is minimal lighting, dark themes work better and
the audiences have a better comfortable amount of light towards their eyes. For very bright-lit environments
light background themes fit well.
Another rule of thumb is to use readable fonts for text and refrain from using stylish ones. Appropriate
sizing and emphasis need to be used at the proper time. It is also a good practice to highlight specific text that
needs to catch the audience’s attention by using appropriate colours for them.
37
1.4 A picture speaks a thousand words
Figure 4: A better slide design without box designs which is more appealing and visually pleasing for audience.
1.4 A picture speaks a thousand words
A common mistake made by presenters is having too much text in their presentations. During a presentation,
the audience’s brain will be largely focused on hearing your words and looking at parts of your slides. The best
practice is to avoid as much as text possible and only include the most required ones. The human brain likes
to see images rather than text. With an image, you can prevent the audience from trying to read through the
text line by line and lose focus on your words. Too much text can cause distractions and the audience will lose
track of the train of thought that you are building. This is very important when you are following the three-part
narrative.
The images should be clearly legible and of high quality and make sure to highlight parts of complicated
ones. For example, say you were to explain the workings of an internal combustion engine to a non-technical
audience. There are two approaches to show this on a slide. One is to write each stage of the work in text with
a single figure with different slides. A better approach is to have a single slide with all the stages and only a title
text as shown in Figure 5.
At times when setting the tone of the presentation, some speakers tend to have just a single image or even
just one text in the middle of the slide. The whole idea of using slide decks is to clearly convey the idea and also
give a visual feast to keep the audience engaged and active throughout the whole time. Generally, it is better to
have more images and fewer text to keep the focus of the audience. Distracting animations and effects should be
avoided that may be gimmicky. Smooth and subtle slide translations and effects are always visually pleasing.
1.5 Voice and Eye contact
Imagine your presentation was like an engaging movie that keeps the audience glued to their seats. A
movie has background music and pleasing visuals to keep you entertained. In a presentation apart from the
slide deck which is the visual part, your sound and eye contact matter the most. Your voice should reflect the
excitement of the story. Appropriate voice modulation where important points are stressed and pausing at the
right places to give a moment of breath changes the mood of one’s presentation. This modulation when followed
by the 3-point narrative, the storytelling will be well harmonized gluing the audience to their seat.
38
1.6 Conclusion
Figure 5: A sample slide to explain the working of an internal combustion engine with a single slide and
minimal text. This helps reduce the number of slides and keeps the focus of the audience.
1.6 Conclusion
It is hard to be a successful scientist without being a good science communicator and similarly for a
businessman without being a great presenter. Communication and presentation skills are key for any career in
science as well as industry. It is the key component for teamwork as well as knowledge dissemination.
Reference
Three Point Act
The Psychology of Color in PowerPoint Presentations
Voice modulation in presentations
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.
39
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.
