SNA Chapter 1 Lecture 3

Summary

This lecture introduces five types of real-world networks: social, biological, information, technological, and language. It explores examples like Twitter and Facebook for social networks, protein interactions for biological networks, and the World Wide Web for information networks. The talk also covers network analysis levels from microscopic to macroscopic, including node properties, community detection, and network motifs. The course aims to teach network properties, dynamics, and applications like node classification, link prediction, and anomaly detection in social media.

Takeaways

• 🌐 There are five main types of real-world networks: social, biological, information, technological, and language networks.
• 🤔 Social networks can be exemplified by platforms like Twitter and Facebook, where users and their relationships form the nodes and links.
• 🧬 Biological networks include protein-protein interaction networks and neural interactions, representing complex biological systems.
• 🌐 Information networks encompass the World Wide Web and citation networks, where nodes represent web pages or documents and links represent hyperlinks or citations.
• 🔌 Technological networks involve infrastructure like power grids, airline, and railway networks, where nodes represent components and links represent connections.
• 💬 Language networks are based on word co-occurrences or keyword relationships within texts, useful for natural language processing tasks.
• 🔍 Network analysis involves studying networks at three levels: microscopic (individual nodes and edges), mesoscopic (clusters or communities), and macroscopic (entire network properties).
• 🔄 Network dynamics include the study of network formation, evolution, and the spread of information or influence within networks.
• 📈 The course will cover applications of network analysis such as node classification, link prediction, and anomaly detection, which are crucial for understanding complex systems.
• 📚 Prerequisites for the course include knowledge of Python programming, probability and statistics, linear algebra, algorithm design, and basics of machine learning and deep learning.
• 🚀 The course aims to provide a comprehensive understanding of network analysis, equipping students with skills to analyze and interpret complex network structures and their applications.

Q & A

What are the five types of real-world networks discussed in the script?

The five types of real-world networks discussed are social networks, biological networks, information networks, technological networks, and language networks.

Can you provide an example of a social network mentioned in the script?

An example of a social network mentioned is the Twitter follower-following network, where nodes are users and links represent the follower-following relationships.

What is a biological network and what are some examples provided in the script?

A biological network refers to interactions within biological systems. Examples include protein-protein interaction networks, neural networks, and food networks.

How is the World Wide Web categorized in the context of network types?

The World Wide Web is categorized as an information network, where nodes can represent web pages and links represent hyperlinks between them.

What are the two types of nodes interaction levels discussed at the microscopic level of network analysis?

The two types of nodes interaction levels discussed are dyadic level, which involves interactions between two nodes, and triadic level, which involves interactions between three nodes.

What does the term 'degree distribution' refer to in network analysis?

Degree distribution refers to the frequency of different degrees (number of connections) present in a network, which is a property analyzed at the macroscopic level.

What is the 'diameter' of a network and how is it determined?

The diameter of a network is the longest shortest path between any pair of nodes in the network. It is determined by measuring the shortest paths between all pairs of nodes and identifying the longest one.

What is a mesoscopic view of a network and what does it involve?

A mesoscopic view of a network involves looking at specific regions or structures within the network, such as clusters or communities where nodes are densely connected, and network motifs which are recurrent sub-structures.

What is the 'small-world property' in networks and how does it relate to the '6 degrees of separation'?

The 'small-world property' refers to the phenomenon where most nodes in a network can be reached from every other node through a small number of steps. The '6 degrees of separation' is a specific example of this property, suggesting that any two individuals are six or fewer acquaintances apart.

How does the script relate social networks to the real world and societal behaviors?

The script suggests that social networks act as a proxy for our society, reflecting public opinion, information consumption patterns, and enabling participation in polls and decision-making processes.

What applications of network analysis are mentioned in the script?

Applications of network analysis mentioned include node classification, link prediction, growth and virality of messages, anomaly detection, and graph representation learning for various purposes like fraud detection and recommender systems.

Outlines

00:00

🌐 Introduction to Real World Networks

The paragraph introduces the concept of real-world networks, emphasizing five main types: social, biological, information, technological, and language networks. Social networks are exemplified by platforms like Twitter and Facebook, where users and their relationships form the nodes and edges. Biological networks include protein-protein interaction networks and neural interactions. Information networks encompass the World Wide Web and citation networks. Technological networks involve infrastructure like power grids and airline routes. Language networks are based on word co-occurrences. The speaker suggests that networks can be abstracted from nearly any dataset, highlighting the versatility of network analysis.

05:00

🔬 Exploring Network Types and Their Structures

This section delves deeper into the types of networks, providing specific examples and discussing their structural nuances. It covers communication networks like wireless mesh networks and the internet, scientific networks such as citation and co-authorship networks, and their directional properties. The importance of understanding clusters or communities within networks is highlighted, along with the concept of language networks, including word co-occurrence networks and keyword co-occurrence networks. The paragraph also introduces semantic networks, akin to knowledge graphs, and their attribute-based connections.

10:05

🏥 Networks in Epidemiology and Multi-Scale Analysis

The paragraph discusses the application of networks in tracking epidemics, like virus spread through patient interactions, and the strategic implications for controlling outbreaks. It introduces three levels of network analysis: microscopic, mesoscopic, and macroscopic. Microscopic analysis focuses on individual nodes and edges, mesoscopic on community structures and network motifs, and macroscopic on the network's global properties like diameter and edge density. The paragraph emphasizes the importance of analyzing networks at different scales to understand their complexity and dynamics.

15:06

🔍 Network Analysis: Properties and Theories

This section ponders whether common properties across different types of networks can lead to a general theory of network structure and dynamics. It mentions properties like the 'small world' effect, 'six degrees of separation', scale-free networks, clustering, robustness against attacks, and cascade effects. The discussion sets the stage for exploring how these properties manifest in various networks and what they reveal about the networks' behavior and resilience.

20:06

📈 Applications and Implications of Network Analysis

The paragraph outlines the applications of network analysis in social media, emphasizing its growth and impact on society. It discusses how social networks are used for public opinion polling and information dissemination. The lecture series will cover various topics including node classification, link prediction, growth and virality of messages, anomaly detection, and graph representation learning. The paragraph also mentions the importance of understanding network-based anomaly detection and the potential of applying network analysis to fraud detection and recommender systems.

25:13

🎓 Course Prerequisites and Conclusion

The final paragraph outlines the prerequisites for the course, recommending knowledge in Python programming, probability and statistics, linear algebra, algorithm design, and basics of machine learning and deep learning. It concludes by encouraging students to learn the skill of networking, indicating the interdisciplinary and applied nature of the course content.

Keywords

💡Social Network

A social network refers to a network of individuals connected through various types of relationships, such as friendships, collaborations, or other forms of social interactions. In the context of the video, social networks are used to illustrate how nodes (users) and edges (relationships) can be represented in a network structure, such as on Twitter or Facebook. The video mentions how these networks can be analyzed to understand patterns of communication and influence.

💡Biological Network

Biological networks are used to describe the complex web of interactions within biological systems. This includes protein-protein interaction networks, where nodes represent proteins and edges represent interactions between them during metabolic processes. The video emphasizes the importance of these networks in understanding biological functions and disease mechanisms.

💡Information Network

Information networks encompass structures that facilitate the flow of information. Examples given in the video include the World Wide Web and citation networks, where nodes are web pages or academic papers, and edges represent hyperlinks or citations. These networks are crucial for understanding how information is disseminated and connected.

💡Technological Network

Technological networks are systems that involve the interconnection of technological components or infrastructures. The video mentions power grids, airline networks, and railway networks as examples. These networks are analyzed to optimize efficiency and reliability, and to understand the impact of failures or disruptions.

💡Language Network

Language networks involve the study of how words and phrases are used and connected in language. The video discusses word co-occurrence networks, where nodes are words and edges represent frequent usage together in sentences. Such networks help in natural language processing tasks like identifying synonyms or detecting language patterns.

💡Network Motifs

Network motifs are recurring patterns of interconnections that appear in complex networks more frequently than in randomized networks. The video uses the example of star and chain motifs to illustrate how these patterns can provide insights into the structure and function of networks, particularly in biological systems.

💡Microscopic Level

The microscopic level of network analysis focuses on individual nodes and edges, examining their properties and interactions. The video explains that at this level, one might look at node properties like degree and centrality, or the interactions between pairs or small groups of nodes, which are fundamental to understanding network behavior.

💡Macroscopic Level

At the macroscopic level, the entire network is considered as a whole. The video describes how properties like the network's diameter, degree distribution, and edge density are analyzed to understand the global structure and characteristics of the network.

💡Mesoscopic Level

The mesoscopic level lies between the microscopic and macroscopic levels, focusing on specific regions or communities within a network. The video mentions how clusters of densely connected nodes can be identified and analyzed to understand local network structures and their roles within the larger network.

💡Small World Property

The small world property is a network characteristic that suggests most nodes can be reached from every other node through a small number of steps or connections. The video introduces this concept to illustrate how networks can be highly interconnected despite their size, which has implications for the spread of information or disease.

💡Scale-Free Property

A scale-free network is one where the distribution of connections among nodes follows a power law, meaning that a few nodes (hubs) have a disproportionately large number of connections. The video discusses how this property influences network resilience and the dynamics of information spread, as seen in social and biological networks.

Highlights

Introduction to the concept of real-world networks and their significance.

Identification of five broad types of networks: social, biological, information, technological, and language.

Examples of social networks, such as Twitter and Facebook, where nodes represent users and links represent relationships.

Explanation of biological networks with protein-protein interaction networks and neural interactions as examples.

Information networks like the World Wide Web and citation networks, highlighting their directed nature.

Technological networks including power grids, airline, and railway networks, emphasizing their critical infrastructure role.

Language networks, including word co-occurrence networks, important for natural language processing.

The idea that networks can be abstracted from almost every complex system or dataset.

Discussion on the microscopic level of network analysis, focusing on nodes and edges.

Macroscopic level analysis, considering the network as a whole, including properties like diameter and edge density.

Mesoscopic view of networks, examining specific regions like communities and network motifs.

Common properties observed across different types of networks, such as small-world property and scale-free property.

Importance of network analysis in understanding the spread of information and misinformation.

Applications of network analysis in anomaly detection, identifying abnormal nodes or activities.

Introduction to graph representation learning (GRL) and its potential in network analysis.

Course content overview, covering network measurement, formation, link analysis, community detection, and more.

Prerequisites for the course, including knowledge of Python, probability, statistics, linear algebra, and basics of machine learning.

Motivational closing, emphasizing the relevance of network analysis in understanding social media and society.

Transcripts

We will discuss different types of  ah you know real world networks;  

I mean the popular real world networks  that we we generally talk about right. So,  

if you if you look at ah the literature um  you can see that broadly there are 5 types  

of networks that we talk about; social network  right, biological network, information network,  

technological network and language network ok.  We will give examples of each of this one by one,  

but in general social networks you all know. Biological networks ah I mean you can think  

of say protein protein interaction  networks or say you know neural network  

I mean interactions between neurons um, food  network and so on. Information networks ah  

include World Wide Web, citation network you know.  Technological network ah net networks include  

power grid, airline network, ah railway network.  Language network includes um you know say  

word co-occurrence network and so on and so  forth. We will discuss each of these one by one.  

In fact, you know I I basically ah tell  my students that you know you can imagine  

you can think of a network from almost every  data set, from almost every application right.  

Network is just an abstraction of a complex  system as I mentioned earlier right. So,  

if you are given a simple problem right ah where  you know a common person cannot see any network  

any notion of network right, you can think  of a network out of it right and that is  

the beauty of this this particular course. So, let us let us look at social network right.  

So, I mean we have been discussing about  Twitter network, multiple times follower  

following network nodes are users and  links can be follower followings and so on.  

Similarly, we have Facebook ah friendship  networks where nodes are nodes ah you know  

users and links are ah friendship relationships  and so on and so forth and this is very obvious.  

We have biological network for example,  protein protein interaction network where  

nodes are you know proteins and two  proteins interact during a different  

metabolic processes in our body and you can  connect proteins accordingly right. Similarly,  

we have you know metabolic network where you know  um basically it describes the relationship between  

you know small say you know metabolites right  and enzymes proteins which basically interact  

with them during different um you know  biochemical reactions for example, right.  

This is a metabolic network and there are tons  of papers on protein protein interaction networks  

mostly in the bioinformatics  computational biology domain.  

You can think of ah other ah communication  network like a ah wireless mesh kind of network,  

where say within a ah within an organization  you can think of different routers computers  

and communications between routers. In fact, you can also think of ah communications  

through satellites that can also be a network  right. Of course, you have a big internet network  

where you know the the the the whole internet is  also considered you know broadly as a network.  

We have scientific networks like a  citation network we discussed earlier where  

papers are nodes and citations are the  links, this is a directed network and  

we have a co-authorship network ah where  nodes are co-authors nodes are authors  

and if two ah authors work together right. You can think of them as co-authors and you  

can connect them right. So, interestingly if  you look at citation network this is a example  

this is a small example of a citation  network. Look at this node right  

look at this node they have they have high  citations, right. In fact, since this is  

a directed network, you have in degree you  have inward edges and outward edges right.  

So, think of a paper which has a lot of inward  edges. So, inward inward edges indicate citations,  

outward edges indicate references right. So, if  a paper has a lot of citations meaning basically  

the paper is very important therefore, people  are citing it, but if a paper say you know  

has a a lot of outgoing edges ah indicating a  lot of references that paper is also important  

that papers might be a book that paper might be a  survey paper or a literature review paper, right.  

We will discuss how this the the the notion of  inward edge and outward edge basically you know  

interplay with each other and you can think of  ah interesting metrics right, out of out of this  

ah the the the notion of directionality  ah of of an edge. Co-authorship network  

if you see the network right you see that you know  there is a closed group, there is another closed  

group here nodes are densely connected right. Here you see yellow nodes densely connected.  

Red nodes are also densely connected right and  and you know every every such group has its own  

identity for example, this red group indicates  researchers working working on agent based models.  

So, green group indicates researchers  working in mathematical ecology  

right, there is this blue group working on  statistical physics and so on and so forth.  

So, you see that a cluster multiple such clusters  basically emerge from a network right which might  

be interesting to study and we will discuss in  a in a separate chapter ah you know how we how  

we can detect such clusters or communities  ah from a network ok. Ah language network:  

one simple example can be a co-occurrence  ah network of words right, where  

nodes can be words and if two words ah co-occur  together in a sentence for example, or very close  

by in a sentence say within within a boundary  within a window of window of 3 words or 4 words  

you basically connect them in the network. So, nodes are words and if two nodes co-occur  

together multiple times you can connect them. For  example, you see that you know the word like um  

teacher right, principal, student they occur ah  very frequently they co-occur very frequently and  

therefore, they are connected right and these kind  of network is very important to study you know ah  

to to automatically detect you know synonymous  words or say antonyms right or say ah holonym  

homonym and so on and so forth, whole bunch of  things in natural language processing right.  

Another example of a network is ah keyword  co-occurrence network where nodes are keywords  

and if two keywords again ah co-occur together  multiple times you can connect them right ah. In  

scientific papers you may have seen that ah you  need to specify keywords right in the paper.  

So, now if you see that you know phrases  like keywords like sentiment analysis and  

opinion mining right. They are very close  and you see that these these keywords  

you know appear together very frequently. So, meaning that you know these phrases are  

very important ah sorry these phrases are  actually you know very linked. You see here  

in this particular network sentiment analysis and  opinion mining they are they are close by right  

and so in this network nodes are keywords and two  keywords are connected if they if they co-occur  

together multiple times in different papers. There is another network called semantic network  

ok. In semantic network now this is basically  knowledge graph you may have heard about the  

term knowledge graph right, a knowledge base or  knowledge graph, where ah nodes are different  

entities indicating different granularities of  knowledge's. For example, you see that cat is a  

mammal right. So, cat is an ah is an entity, this  is a node mammal is a node right ah ah whale is a  

node, animal is a node and so on and so forth. So, cat is a mammal. So, therefore, there is a  

link from cat to mammal ah and the relationship  is a right. So, this is you can think of this  

as an attributed network where edges are  associated with some sort of attributes  

right. Is a or has right and so on and so  forth leaps in and so on and so forth right,  

these are different attributes of edges right.  So, now this is called semantic network ok.  

Similarly, we have we can think of many other  interesting networks like ah you know terrorist  

network. I mentioned last day that ah you know  ah where in this particular network nodes can be  

terrorists and if two terrorists went together  for a similar mission or if two terrorists  

were arrested together right you can basically  connect them through ah through links right ah.  

In fact, there are very interesting studies  ah where people basically keep track of how  

such movement happens. I mean I am not very  aware of this, but if you look at studies  

there are multiple books on you know how you  can model such activities using using networks  

ok. Ah There is another type of network called  patient network and this is very important in  

again in the computational biology ah domain  where you basically want to study how a particular  

epidemic spreads or ah or a ah virus spreads. So, here nodes are patients and if two patients  

are interacting if two patients come  come closer together you can connect them  

right. So, now if say let us let us assume that  you have ah in a hospital you have this kind of  

network right and you have certainly seen  that a virus has started spreading right.  

So, you you will immediately understand  that through which path this virus has  

basically spread because you know that  these people have been infected right  

and you also know that these people have have  been frequently interacting with other people  

right. So, essentially it means that you may  want to protect those ah people who have already  

you know been interacting with the ah already  infected patients and you want to protect them  

either through vaccinations or some other ways. So, now we will ah look at a network from  

different angles right. We will basically inspect  ah the network as a whole we then zoom in right  

and try to ah look at a a part of the network.  We further zoom in and look at you know ah even  

more fine grained entities in a network right.  So, we will discuss three levels of granularity  

of a network, one is called microscopic level, the  other is called macroscopic level. So, you have  

microscopic level you have macroscopic level  of analysis of network and in between you have  

something called mesoscopic level  mesoscopic view of a network ok.  

So, let us start with microscopic level. So, when  you talk about microscopic microscopic view of a  

network we basically ah you know we basically  analyse nodes and edges ok. We look at nodes  

properties, we look at edges properties, we look  at how two nodes interact right, how say three  

nodes interact and so on. We do not go further  ok. So, we we look at different properties of  

nodes for example, degree, centrality we will  discuss what is centrality later right and so on.  

We look at how two nodes interact right and  this is called a dyadic level of interaction.  

We will look at how three nodes interact  this is called triadic level of interaction  

right. For example, say this is a triadic level  of interaction or this is another type of triadic  

level of interaction right. This is another  type of triadic level of interaction and so on.  

So, we look at dyadic dyadic level of  interaction, triadic level of interaction  

and we also look at egocentric circle. We have already discussed what is ego  

net ah ego network right. We will see that  say let us say this is a ego network ok.  

And let us say this is a structure and this is  ego, this is ego and these are the altars ok. So,  

you see that here there is a circle ok meaning  a closely connected nodes right. This is also  

a closely connected ah ah group ok. So, these are  called circles egocentric circles. This is also a  

kind of a microscopic level view because you  are only looking at a particular node and  

its surrounding neighbours and that is all ok. Now, let us look at macroscopic level the entire  

network as a whole right. We can think of you know  number of nodes, degree distribution ok this the  

term degree distribution may not be very familiar  to you, but we will discuss in the next chapter.  

There is something called the diameter  of a of a network which is the you know  

longest shortest path right. You may have  heard about something called shortest path,  

you take a pair of nodes and look at the  shortage path you take all pairs of nodes right.  

You can see which one is the longest right. So,  this is the diameter, that is called the diameter  

of a network, it is a network property. Similarly  you can think of edge density of a network right.  

How many edges can be formed? How many edges  can be possible ah in a network of ah node n  

n number of nodes n c 2? And how many edges  are actually there in the network right?  

So, you take a fraction of actual number of edges  divided by ah I mean the fraction of actual number  

of edges and the possible number of edges and  that will give you something called edge density  

right. So, this is basically ah looking at the  network as a whole micro macroscopic level.  

In between microscopic and macroscopic there is  something called mesoscopic view of a network  

ok. And what is this? Mesoscopic view we we look  at specific regions of a network. For example  

I already mentioned about something  called clusters or communities  

where multiple nodes interact together frequently  and form a dense group right, you can think of  

this as a mesoscopic structure of a network. You have some multiple such communities and  

these communities then form the entire network  right. You can also think of something called  

network motifs and this motif structure is very  useful in in biological network particularly,  

where now what is motif? Motif is basically  a recurrent you know sub structures  

ah which appear in a network right. For example,  if you think of this [FL] this is a star network  

right and this star network appears very  very frequently in a in a network. So,  

that network gives you a separate indication. If you see a chain this is a chain right this  

chain the the the this is called chain motif  right. Now if you see a lot of such chain motifs  

present in a network the that network  can be different from a network having  

a lot of star motifs for example, right. So,  again motif analysis is a different ah you  

know different ah direction of network  network analysis in general which we are  

not covering because this is more related to  biological ah network. So, but but of course,  

in social network also we have studied we we we  see plenty of cases where motifs are useful.  

Now, you know this if you look at the science of  network analysis. So, the question that we ask  

is that are these properties common across  networks? Let us say you take a a Facebook network  

and you take a protein protein interaction  network right and you see that some of the  

properties are some of the microscopic mesoscopic  macroscopic properties are common across networks,  

then what do you conclude right? Would  you be able to conclude that these two  

networks are same or have similar properties?  What are the different properties right?  

So, the question that we ask is that can  we formulate a general ah theory of the  

structure of the structure evolution  and the dynamics of a network ok. Ah  

The common the the the the the observed property  the common observed property ah that we often see  

in a network includes something called small world  property ok. We will discuss in the next chapter  

what is small world property. It basically says  that you know the world is very small meaning that  

if you want to move from one node to another  node you do not need to traverse a lot ok.  

And there is a very interesting property called  6 degree of separation ok. It basically says  

that you know ah there are 6 hoops on an average  between pair any pairs of ah any pair of nodes  

in the network ok. We will also discuss something  called scale free property right. We will discuss  

clustering community structure we will discuss  something called the robustness of a network  

robustness of a network to different  attacks different adversarial attacks.  

We will discuss something  called cascade effect ok.  

The vulnerability ah to different cascading  failure. For example, say there is a power grid  

electric power grid and suddenly you see that  one one ah node ah has got damaged right. And  

if one node got damaged it it may happen  that this damage may get propagated through  

the entire network right entire power grid, but  that that that will you know create devastation  

right. So, we will need to stop such ah such  power failure, such failure of nodes right.  

So, what would be the strategies through which we  can we can stop the spread of such such failure?  

So, this course will cover all these things in  different chapters and we will basically learn  

different mathematical formulations properties  which can define ah this this characteristics of  

of a network. Now this is a motivational slide  I am pretty sure that you do not need to ah I  

do not need to motivate you why you know social  media has become a part and parcel of for life.  

If you see the user engagement over time right  right from say the year 2000 to 2021 2022  

there is a massive growth exponential growth of  usage particularly during this you know lockdown  

time ah this pandemic time right and a lot of  content are being generated every minute right.  

So, people sometimes say that you know  social network is a proxy of our society  

right. You see many cases where you know people  take ah opinion ah public poll right. You you you  

have you know may have been ah invited to ah to to  to vote for certain certain polls ah right certain  

ah ah certain decision right and you know on the  in the in the online social media ah since ah  

you you are observing the patterns you  are observing the ah different opinions.  

You are consuming different  information in different ways  

right. You are one of the stakeholders right  who can participate in this polling right. So,  

therefore, this is this is very important ok. So,  these are the applications that we will also cover  

in this in this particular lecture, we will cover  something called node classification where we  

classify nodes depending upon the properties. We predict links right this is also called  

recommendation link recommendation or and and link  recommendation has a lot of applications in the  

product recommendation friendship recommendation  ad recommendation and so on. We will discuss  

something called a growth and virality of of  messages right of of networks in general.  

We will discuss node centric network centric  properties personalized properties of nodes.  

We discuss how misinformation spreads over  social network, we discuss you know ah how  

you can identify nodes which are abnormal for  example, say fraud nodes right fraud users  

and there is something called anomaly detection  through which we ah look at how we can identify  

such ah you know fraud nodes or outlier nodes. Now, this is the tentative course content  

that we are going to cover in this  lecture series. We will start by ah  

you know by measuring a network. We quantify  in the next lecture we will try to quantify  

a network, then we will something we will discuss  something called network network formation. We  

will discuss many such models through which you  can actually mimic the way a network is formed,  

a way a network ah evolves over time right. We will discuss ah you know random growth  

model ah, we will discuss preferential achievement  models and many such models those models are  

mostly borrowed from physics, but they are highly  applicable here in social network as well. We will  

discuss link analysis where you specifically look  at how to characterize a link and edge right and  

how we will see how different social theories are  basically ah are useful to characterize a link.  

We will discuss community detection  community or cluster is a very important  

property of a network and we will see how  we can detect communities efficiently.  

We discussed something called link prediction ok  and we have already mentioned the application in  

recommendation system, but the problem is you know  when we need to predict something for ah future  

ah given that we have network ah ah you know at  the current time stamp this is difficult because  

you do not know what is going to happen right. So, link prediction is very very challenging and  

we will discuss some ah you know some algorithms  which have become very popular because of  

because of their you know the the way they are  

they basically work. We discussed ah network  effect and cascade behaviour particularly ah how  

this epidemic spreads over over network  online and offline network. We discussed  

network based anomaly detection ah  anomaly detection or abnormality detection  

ah is also studied in data mining. But, when it comes to network structure ah  

approaches are very different and we will see ah  how ah we ah identify anomalous nodes anomalous  

entities ah it can be anomalous edges. It can be  anomalous sub graphs right ah from the network.  

Then we will discuss a very interesting in fact um  ah in fact, ah this is a recent trend in network  

analysis something called graph representation  learning network representation learning GRL ok  

in short. So, we will see how network is mapped,  how a network is mapped to an embedding space  

right to an Euclidean space for example, right. And when you map a network ah to a vector space  

say on an Euclidean space things would become very  easy. For example, now a node is represented by a  

vector right. So, you can do whole bunch of  vector operations matrix multiplication and  

so on and so forth to solve ah ah different  applications, but to understand this chapter  

you need to ah know a basics of ah deep learning,  basics of ah machine learning we do not need to go  

into details of that, but a basics of machine  learning and deep learning might be useful.  

And then we will conclude this ah lecture by  you know by by giving you ideas about some  

applications for example, fraud detection. We  look at fraud detection particularly right,  

we will we look at something called collusion ah  black market driven activities in online social  

network. How you can ah detect such activities,  we will also discuss very briefly about  

recommender systems like particularly friendship  ah recommendation system and so on right ah.  

So, these are the prerequisites. So, we so  ah it is highly recommended that you learn  

python programming ah, you I assume that ah I am  assuming that you have fair bit of ideas about  

probabilities and statistics I mean probability  statistics 101 is good enough. Again I am assuming  

that you have ideas ah basic ideas about  linear algebra particularly matrix operations  

ah you have ideas about ah basic algorithm design  right and ah it would be great if you also learn  

basics of machine learning and deep learning ok. With this ah I would ah like all of you to ah  

learn together the skill of networking. Thank you.