Machine Learning in Social Networks

Embedding Nodes, Edges, Communities, and Graphs

Authors: Manasvi Aggarwal, Prof. M.N. Murty

Publisher: Springer Singapore

Book Series : SpringerBriefs in Applied Sciences and Technology

Part of: Springer Professional "Business + Economics & Engineering + Technology" , Springer Professional "Engineering + Technology" , Springer Professional "Business + Economics"

Table of Contents

About this book

This book deals with network representation learning. It deals with embedding nodes, edges, subgraphs and graphs. There is a growing interest in understanding complex systems in different domains including health, education, agriculture and transportation. Such complex systems are analyzed by modeling, using networks that are aptly called complex networks. Networks are becoming ubiquitous as they can represent many real-world relational data, for instance, information networks, molecular structures, telecommunication networks and protein–protein interaction networks. Analysis of these networks provides advantages in many fields such as recommendation (recommending friends in a social network), biological field (deducing connections between proteins for treating new diseases) and community detection (grouping users of a social network according to their interests) by leveraging the latent information of networks. An active and important area of current interest is to come out with algorithms that learn features by embedding nodes or (sub)graphs into a vector space. These tasks come under the broad umbrella of representation learning. A representation learning model learns a mapping function that transforms the graphs' structure information to a low-/high-dimension vector space maintaining all the relevant properties.

Frontmatter

Chapter 1. Introduction

Abstract

Networks can be used to abstract most real world applications because of which network analysis has become an interesting field of research. But networks are high-dimensional raw data structures and therefore can challenge the analysis tools in terms of both computational and performance metrics. This prompted recent growth in network embedding tools and techniques to present the underlying data in a simpler form for analysis. In this chapter the notion of embedding is introduced. Also how an embedding helps in overcoming some of the problems associated with the analysis of large high-dimensional networks.

Manasvi Aggarwal, M. N. Murty

Chapter 2. Representations of Networks

Abstract

Networks play a pivotal role in representing relational data. Network analysis is gaining importance because of its relevance to several real-life applications. We deal with an introduction to social and information networks and their representations in this chapter. We introduce network embeddings followed by Matrix Factorization approaches. Further, we examine related datasets, evaluation metrics, and the downstream machine learning tasks. An introduction to one of the most influential developments in the recent times, word2vec and its role in embeddings, and categories of embeddings is also provided.

Manasvi Aggarwal, M. N. Murty

Chapter 3. Deep Learning

Abstract

This chapter deals with a brief introduction to deep learning. We deal with the perceptron classifier and its training. We then deal with feedforward networks and the multilayer perceptron (MLP). Training MLP using the well-known backpropagation algorithm is examined. An introduction to convolutional neural networks (CNNs), recurrent neural networks (RNNs), Long Short-TermMemory (LSTM), and autoencoders is provided.

Manasvi Aggarwal, M. N. Murty

Chapter 4. Node Representations

Abstract

Downstream ML tasks can exploit the low-dimensional node embeddings by using them as the inputs to traditional machine learning models. These node-level downstream tasks include node classification, node clustering, recommendation, link prediction, and visualization. In this chapter, we discuss node embedding techniques. These techniques are based on one of random walk, matrix factorization, or deep learning. Further, some algorithms learn representations in an unsupervised setting while others learn in a supervised setting. We finally present comparison of these algorithms according to their performance on downstream tasks.

Manasvi Aggarwal, M. N. Murty

Chapter 5. Embedding Graphs

Abstract

There are several applications where an embedding or a low-dimensional representation of the entire graph is required. This chapter deals with such representations which are called graph embeddings. We consider various state-of-the-art graph pooling techniques that are important in this context. We also consider graph level analysis tasks including graph classification, and graph visualization. We also compare them using several benchmark evaluation datasets.

Manasvi Aggarwal, M. N. Murty

Chapter 6. Conclusions

Abstract

this book we have examined social and information networks, and their analysis. Specifically, we considered the following aspects.

A fundamental problem in data analysis is representation. So, representation learning is the most important step in dealing with almost any large-scale practical problem.

In this book we have examined in detail different schemes for network representation learning (NRL).

There was more emphasis on social and information networks in the book. However, the schemes discussed are generic and can be applied to any other complex network.

It is important to note that data in the form of networks is either explicit or implicit where the networks are typically represented as graphs.

The importance of networks in dealing with any application need not be over emphasized. They are so important that the schemes considered in the book are useful in both implicit and explicit cases.

The basic problem examined in detail in the book is embedding network entities. Both node and graph embedding schemes are examined in detail. Further, state-of-the-art embedding schemes are compared using several benchmark datasets.

The background required in terms of graphs, adjacency matrices, matrix factorization, random walks, representing words as vectors, neural networks, and deep learning schemes are discussed in detail in Chaps. 2 and 3.

Evaluation of various embedding schemes is typically done with the help of downstream ML tasks including classification, community detection, link prediction and visualization. We have explained these ML tasks in Chap. 2.

Different schemes for embedding nodes in a network are examined in Chap. 4. In Chap. 5, various schemes for embedding an entire graph are considered.

10.

A brief summary of the importance of networks and their representations is done in the current chapter with a view that networks will play an important role, in every practical application, in the near future.

Manasvi Aggarwal, M. N. Murty

Backmatter

Title: Machine Learning in Social Networks
Authors: Manasvi Aggarwal
Prof. M.N. Murty
Publisher: Springer Singapore
Electronic ISBN: 978-981-334-022-0
Print ISBN: 978-981-334-021-3
DOI: https://doi.org/10.1007/978-981-33-4022-0