Hot Topics in Machine Learning for

Research and Thesis

Machine Learning and its subsequent fields have undergone tremendous growth in the past few
years. It has a number of potential applications and is being used in different fields. A lot of
research work is going on in this field. There has been a lot of buzz around this field in the recent
times. It is the major application of Artificial Intelligence. Algorithms are a major component of
Machine Learning. One should have a complete understanding of these algorithms before doing
research on different topics in Machine Learning. There are various topics in Machine Learning
for M.Tech thesis and Ph.D. research.

Here is the list of hot topics in Machine Learning for thesis and research:

 Deep Learning
 Human-computer interaction
 Genetic Algorithm
 Image Annotation
 Reinforcement Learning
 Natural Language Processing
 Supervised Learning
 Unsupervised Learning
 Support Vector Machines(SVMs)
 Sentiment Analysis

Deep Learning

Deep Learning is a sub-field of Machine Learning or we can say it is an advanced version of

Machine Learning. Deep Learning can also be referred to as deep structure learning or
hierarchical learning. It is one of the hot topics in machine learning for master's thesis and
research. The concept of deep learning is being used by big companies like Google, Amazon to
increase their productivity and sale rate.

The algorithms in deep learning or deep neural networks are concerned with the functioning of
the human brain and its structure. Deep Neural Network is a type of neural network having more
than two layers. This type of neural network needs more data as well as the computational power
to derive results.
Applications of Deep Learning

Deep Learning applications will significantly affect our daily life in near future. Some of the
applications have already made their impact. Here are some of the important applications of deep
learning:

 Image Recognition
 Voice Assistants
 Self-driving cars
 Computer-aided medical diagnosis
 Automatic Machine Translation

Limitations of Deep Learning

There are some limitations of deep learning which are as follows:

 It needs a large amount of data to extract results.

 Substantial computational power and resources are required by deep neural networks.
 Deep Learning is a time-consuming process.
 A training is to be provided so as to enable deep learning to make decisions.
 A high-performance computing environment is required for deep learning.

Human-computer Interaction

Human-computer interaction or HCI is the study of human and computer activities and how they
interact with each other. It is a very good field for research in machine learning. There are
different ways in which humans interact with computers and HCI deals with the study of this
interaction. To facilitate this interaction, an interface is required between humans and computers.
A graphical user interface is one such example of the interface used by desktop applications and
internet browsers. Similarly, voice user interfaces(VUI) are used for speech recognition.

The idea of HCI dates back to early 1980s. It is a very broad field covering the areas like user-
centered design, user experience design, and user interface design. Research work is going on the
following areas of HCI:

 Augmented Reality
 Social Computing
 Brain-computer interface
 User Customization
 Embedded Computation

Genetic Algorithm

The concept of Genetic Algorithm is based on the principle of Genetics and Natural Selection
and is a search-based optimization technique used to find optimal solutions to complex problems.
It is another good topic in machine learning for thesis and research. It is the most efficient tool to
solve difficult problems referred to as NP-Hard problems.

Genetic Algorithms are important in machine learning and are based on the following three types
of rules:

 Selection rules to select the parents from the current population

 Crossover rules to combine two parents to produce children for the upcoming generation
 Mutation rules to apply changes to parents to produce children
Applications of Genetic Algorithm

Following are some of the applications of Genetic Algorithm:

 Automotive Design
 Robotics
 Encryption
 Computer-Aided Design
 Bioinformatics
 Machine Learning Feature Selection
 Mutation Testing
 Software Engineering

Image Annotation

Image Annotation is a process in which a caption or keyword is assigned to a digital image

automatically. It finds its application in image retrieval systems to locate images from the
database. Machine Learning methods and algorithms are applied to Automatic Image
Annotation. Clustering and classification are the most commonly used methods in the process of
image annotation.

Image Annotation Tools

There are various tools for manual image annotation some of which are listed below:
  DataTurks
 Labelbox
 AnnoStation
 LabelMe
 Pixorize
 Microsoft VoTT
 Images Annotation Programme
 FastAnnotationTool

Reinforcement Learning

Reinforcement Learning is a type of machine learning algorithm in which an agent learns how to
behave in an environment by interacting with that environment. A lot of research has been done
in this area of machine learning in the recent times. It mostly finds its application in gaming and
robotics. The approach of this algorithm is different from other machine learning algorithms
which are supervised learning and unsupervised learning.

The definition of reinforcement learning can be understood with the following concepts:

 Agent - An agent is the one that takes action in an environment.

 Action(A) - It is the series of steps taken by an agent in an environment.
 Environment - The real world in which the agent takes an action.
 State(S) - The situation of an agent at any particular time.
 Reward(R) - A type of feedback through which the success and failure of user's actions
are measured.

Natural Language Processing

Natural Language Processing or NLP is a branch of Artificial Intelligence using which

computers are made to understand, manipulate, and interpret human language. It aims to fill the
space between human communication and computer understanding. It is another good topic in
machine learning for thesis and research. It uses the concept of machine learning and deep
learning for complete interaction between humans and computers.

Importance of Natural Language Processing

The importance of natural language processing lies in the fact that it enables computers to
communicate with humans in their own language. Computers can interpret human speech and
text using the concept of natural language processing. It will help to analyze the large volumes of
textual data generated every day.
Applications of Natural Language Processing

Following are the main applications of Natural Language Processing:

 Speech Recognition
 Language Translation
 Caption Generation
 Language Modelling
 Optical Character Recognition(OCR)
 Information Retrieval
 Question Answering
 Sentiment Analysis
 Text Segmentation
 Document Clustering

Supervised Learning

It is a type of machine learning algorithm in which both input and output data is provided and the
output data is mapped to the input through a mapping function. In other words, supervised
learning is a type of machine learning algorithm that uses training datasets for making decisions.
There are two types of algorithms in supervised learning which are:
  Classification - Where data can be categorized into specific classes for categorical
response values. Commonly used classification algorithms are:
o Support Vector Machines(SVM)
o Neural Networks
o Decision Trees
o Discriminant Analysis
o Nearest neighbors(kNN)
 Regression - Regression algorithms are used for continuous-response values. Following
algorithms are included in this category:
o Linear Regression
o Nonlinear Regression
o Generalized linear models
o Decision Trees
o Neural Networks

Supervised learning has a number of applications including algorithm trading, credit scoring,
tumor detection, drug discovery, pattern recognition, price forecasting to name a few. It is also a
very good thesis topic in machine learning.

Unsupervised Learning

Unsupervised Learning is a type of machine learning algorithm to find hidden patterns and
underlying data structures. The inferences are drawn by this algorithm from the datasets
containing the input data. Cluster Analysis is the most commonly used method in unsupervised
learning. General clustering algorithms in unsupervised learning are:

 Hierarchical Clustering
 k-Means Clustering
 Self-organizing maps
 Hidden Markov Models
 Gaussian mixture models

The main purpose of unsupervised learning is to group data having similar characteristics into
different clusters. This done by finding datasets having similar patterns. Apart from clustering,
Principal Component Analysis(PCA) is another commonly used technique in unsupervised
learning. Unsupervised Learning finds its application in data mining, text mining, bioinformatics,
image segmentation, computer vision, and genetic clustering.

Support Vector Machines(SVMs)

Support Vector Machines or SVMs are one of the most important machine learning algorithms.
The purpose of this algorithm is to analyze the data used for classification and regression
analysis. As compared to other algorithms, the concepts of SVM are relatively simple. Using
kernel trick, SVMs can perform non-linear classification. In this algorithm, each data item is
plotted as a point in n-dimensional space where n is the number of features. After that,
classification is performed on data items. A hyperplane will be found that will divide the datasets
into two different classes.

Advantages of Support Vector Machines

1. It provides accurate results

2. It works really well on smaller datasets
3. It is one of the most efficient machine learning algorithms

Drawbacks of Support Vector Machines

1. Not suitable for larger datasets as training time is high

2. Discrete data is another drawback of SVMs

Applications of Support Vector Machines

SVMs have various applications some of which are:

 Image Classification
 Face Detection
 Bioinformatics
 Text Categorization
 Handwriting Recognition
 Protein Remote Homology Detection

Sentiment Analysis

Sentiment Analysis is also known as opinion mining and is a process to determine whether the
attitude of an individual towards a product or topic is positive, negative 0r neutral expressed in
the form of text. It is another good topic in machine learning for thesis and research. It uses the
concept of natural language processing, machine learning, computational linguistics, and
bioinformatics to extract essential information. It is mainly used in case of social media
monitoring. Sentiment Analysis is crucial such that it helps to find what a customer thinks of a
particular brand.
Sentiment Analysis Tools

Here are some tools which help to track user sentiments:

 Google Analytics
 Hootsuite
 Pagelever
 Marketing Grader
 Facebook Insights
 Google Alerts
 Meltwater
 Tweetstats

Applications of Sentiment Analysis

Sentiment Analysis can be used in different areas for different purposes. Following are some of
the applications of sentiment analysis:

 Online Commerce
 Voice of the Customer(VOC)
 Voice of the Market(VOM)
  Brand Reputation Management
 Voting advise applications
 Government Intelligence

These are the hot topics in Machine Learning for thesis and research although there are various
other topics also. Machine Learning is one of the trending fields for the thesis in computer
science.