Machine Learning and Deep Learning Algorithms

Tools and Techniques Using MATLAB and Python

Abhishek Kumar Pandey, Abhishek Kumar Pandey, Pramod Singh Rathore, Dr. S. Balamurugan

• Publisher: BPB Publications
• Year: 2019
• eBook ISBN: 9789388511131
• Topic: Computer Science
• Subtopic: Computer Science General
• Index: Computer Science

Guide covering topics from machine learning, regression models, neural network to tensor flow Key Features

• Machine learning in MATLAB using basic concepts and algorithms.
• Deriving and accessing of data in MATLAB and next, pre-processing and preparation of data.
• Machine learning workflow for health monitoring.
• The neural network domain and implementation in MATLAB with explicit explanation of code and results.
• How predictive model can be improved using MATLAB?
• MATLAB code for an algorithm implementation, rather than for mathematical formula.
• Machine learning workflow for health monitoring.

Description
Machine learning is mostly sought in the research field and has become an integral part of many research projects nowadays including commercial applications, as well as academic research. Application of machine learning ranges from finding friends on social networking sites to medical diagnosis and even satellite processing. In this book, we have made an honest effort to make the concepts of machine learning easy and give basic programs in MATLAB right from the installation part. Although the real-time application of machine learning is endless, however, the basic concepts and algorithms are discussed using MATLAB language so that not only graduation students but also researchers are benefitted from it. What Will You Learn

• Pre-requisites to machine learning
• Finding natural patterns in data
• Building classification methods
• Data pre-processing in Python
• Building regression models
• Creating neural networks
• Deep learning

Who This Book Is For
The book is basically meant for graduate and research students who find the algorithms of machine learning difficult to implement. We have touched all basic algorithms of machine learning in detail with a practical approach. Primarily, beginners will find this book more effective as the chapters are subdivided in a manner that they find the building and implementation of algorithms in MATLAB interesting and easy at the same time. Table of Contents

• Pre-requisite to Machine Learning
• An introduction to Machine Learning
• Finding Natural Patterns in Data
• Building Classification Methods
• Data Pre-Processing in Python
• Building Regression Models
• Creating Neural Networks
• Introduction to Deep Learning
• 
  About the Author
  Abhishek Kumar Pandey is pursuing his Doctorate in computer science and done M.Tech in Computer Sci. & Engineering. He has been working as an Assistant professor of Computer Science at Aryabhatt Engineering College and Research center, Ajmer and also visiting faculty in Government University MDS Ajmer. He has total Academic teaching experience of more than eight years with more than 50 publications in reputed National and International Journals. His research area includes- Artificial intelligence, Image processing, Computer Vision, Data Mining, Machine Learning. He has been in International Conference Committee of many International conferences. He has been the reviewer for IEEE and Inder science Journal. He is also member of various National and International professional societies in the field of engineering & research like Member of IAENG (International Association of Engineers), Associate Member of IRED (Institute of Research Engineers and Doctors), Associate Member of IAIP (International Association of Innovation Professionals), Member of ICSES (International Computer Science and Engineering Society), Life Member of ISRD (International Society for research & Development), Member of ISOC (Internet Society).He has got Sir CV Raman life time achievement national award for 2018 in young researcher and faculty Category. He is serving as an Associate Editor of Global Journal on Innovation, Opportunities and Challenges in Applied Artificial Intelligence and Machine Learning. Blog: http://veenapandey.simplesite.com/
  LinkedIn Profile: https://www.linkedin.com/in/abhishek-pandey-ba6a6a64/ Pramod Singh Rathore is M. Tech in Computer Sci. and Engineering from Government Engineering College Ajmer, Rajasthan Technical University, Kota, India. He have been working as an Assistant Professor Computer Science at Aryabhatt Engineering College and Research center, Ajmer and also a visiting faculty in Government University Ajmer. He has authored a book in Network simulation which published worldwide. He has a total academic teaching experience more than 7 years with many publications in reputed national group, CRC USA, and has 40 publications as Research papers and Chapters in reputed National and International E-SCI SCOPUS. His research area includes machine learning, NS2, Computer Network, Mining, and DBMS. He has been serving in editorial and advisory committee of Global journal group, Eureka Group of Journals.He has been member of various National and International professional societies in the field of engineering & research like Member of IAENG (International Association of Engineers). Dr S. Balamurugan is the Head of Research and Development, Quants IS & CS, India. Formely, he was the Director of Research and Development at Mindnotix Technologies, India. He has authored/co-authored 33 books and has 200 publications in various international journals and conferences to his credit. He was awarded with Three Post-Doctoral Degrees- Doctor of Science (D.Sc.) degree and Two Doctor of Letters(D.Litt) degrees for his significant contribution to research and development in Engineering, and is the recepient of thee Best Director Award, 2018. His biography is listed in "World Book of Researchers" 2018, Oxford, UK and in "Marquis WHO'S WHO" 2018 issue, New Jersey, USA. He carried out a healthcare consultancy project for VGM Hospitals between 2013 and 2016, and his current research projects include "Women Empowerment using IoT", "Health-Aware Smart Chair", "Advanced Brain Simulators for Assisting Physiological Medicine", "Designing Novel Health Bands" and "IoT -based Devices for Assisting Elderly People". His professional activities include roles as Associate Editor, editorial board member and/or reviewer for more than 100 international journals and conferences. He has been an invited as Chief Guest/Resource Person/Keynote Plenary Speaker in many reputed Universities and Colleges His research interests include Augmented Reality, the Internet of Things, Big Data Analytics, Cloud Computing, and Wearable Computing. He is a life member of the ACM, ISTE and CSI LinkedIn Profile: https://www.linkedin.com/in/dr-s-balamurugan-008a7512/
  

CHAPTER 1

An Introduction to Machine Learning

Machine learning belongs to a sub class field of a class of Artificial Intelligence (AI). Machine learning performs effectively by considering an effective information structure in order to appropriately fit the information in models for individual comprehension.

In the field of software engineering, machine learning algorithms exhibit a desire performance due to computational methodologies. PC issues which are not customized are usually calculated customary with ascertain factors. In these cases, machine learning performs a calculation in computers by considering factual examination for estimating a particular information range. Further to this machine learning motivates the for input information processing machine learning encourages test information in robotize of information inputs.

Many innovative techniques have been derived through machine learning, including web-based social networking that provides tagging, photographs, and social networking factors. On the other hand, the optical character acknowledgment (OCR) facilitates changes in the pictures of mobile. General techniques derived from machine learning enable motion pictures for TV shows based on user inclinations. To make ease off buyers, self-driving autos are explored for figuring out machine information factors.

The field of machine learning exhibits a constant progress for desired performance with desired field creation. General classification of machine learning techniques offer an ordered part undertaking. This classification in machine learning is accomplished by the framework that is adopted and created in the network for machine learning.

Most commonly adopted techniques by large grasped machine learning procedures—supervised learning, trains counts in light of representation data and yield data that is set apart by individuals. Unsupervised learning does not provide the computation with named data wherein remembering the true objective to empower it is to find structure inside its data.

1.1 Basics of Machine Learning

Machine learning has been around for a long time now and every single social media client, have at some point in time been customers of machine learning innovation. One of the basic illustrations is confront acknowledgment programming, which is the ability to distinguish whether a computerized photo incorporates a given individual.

Definition

“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.”

1.2 Machine Learning Types

Machine learning is of two types:

• Supervised learning
• Un-supervised learning

Supervised Learning

Machine learning algorithm has a major functionality of utilization of learning techniques through machines. In supervised approach, we are able to get the output variable data (Y) through the input variable (X) for learning mapping function.

Y = f(X)

The primary objective here is to map function approximation for newly defined input data (X) and for output data prediction variables (Y) for a given data.

Supervised learning is obtained through training a dataset which is similar to the learning process via supervision of a teacher. We are already aware of the correct answer, through iterative prediction of training the data and rectify the process through teaching. In this learning, we will be stopped when desired acceptable performance is achieved in the machine.

Even supervised learning is subject to classification and regression as major problems for performance measurement. Let’s now understand classification and regression in more detail:

• Classification: For this particular problem category, the output variable is classified as disease, no disease, blue, or red.
• Regression: When the output variable is obtained as a real value, regression problem will occur. For example: weight or dollars.

A few problems that are built with classification and regression require time series prediction and recommendation respectively. The most popular supervised machine learning algorithms are stated with problem solutions:

• Linear regression to solve regression problems
• Random forest to solve regression and classification problems
• Support vector machines to solve classification problems

Unsupervised Learning

If the data only consists of input data (X) without any output variables corresponding to the input variables, then, it belongs to unsupervised machine learning algorithm.

The primary goal of unsupervised learning is data distribution with underlying structure for more understanding of data.

This kind of data processing is known as unsupervised learning since it does not have the correct answers or a teacher for learning as in the case of supervised learning. In this, unsupervised learning algorithms are left in own devices for discovering and structure intersecting in the data.

Similar to supervised learning unsupervised learning are also grouped as association and clustering problems, let’s understand what these are:

• Clustering: Clustering problem arises for discovering data with inherent groups similar to purchasing behavior of customer groups.
• Association: Learning problem with association rule is to discover rules for data at a larger portion just like people tend to buy the variable product X and also buy variable Y.

A few examples for unsupervised learning algorithm are:

• Clustering problem is resolved with the development of k-means
• To resolve association rule learning problem Apriori algorithm is developed

Approaches

Machine learning is immovably related to computational bits of knowledge. Hence, having an established data in estimations is significant for cognizance and using machine learning algorithms. It can be helpful to first describe association and backslide as they are routinely used methodologies for investigating the relationship among quantitative variables. Correlation is a measure of connection between two factors that are not allocated as either poor or independent. Regression at a key level is used to have a glance at the association between one dependent and one self-ruling variable. Since backslide bits of knowledge can be used to presume the dependent variable when the self-ruling variable is known, backslide engages desire capacities.

K-nearest neighbor

The K-nearest neighbor calculation is an example acknowledgment show that can be utilized for arrangement and in addition relapse. Frequently curtailed as k-NN, the k in k-nearest neighbor is a positive whole number, which is normally small. In either order or relapse, the information will comprise of the k-nearest preparing cases inside a space.

We will center around k-NN grouping. In this strategy, the yield is class participation. This will lead to another question—which class is the most basic one amongst its k-nearest neighbors? On account of k = 1, the question is doled out to the class of the single closest neighbor.

When we pick k = 3, the calculation will locate the three closest neighbors of the green heart to group it to either the precious stone class or the star class.

In our outline, the three closest neighbors of the green heart are one jewel and two stars. Therefore, the calculation will arrange the heart with the star class.

Among the fundamental machine learning calculations, k-nearest neighbor is thought to be a kind of apathetic learning as the speculation past the preparation information does not happen until an inquiry is made to the framework.

Decision Tree Learning

For general use, decision trees are used to apparently address decisions and show up or light up essential authority. When working with machine learning and data mining, decision trees are used as a judicious models. These models depict about data to choices about the data goal regard.

The target of decision tree learning is to influence a model that will foresee the estimation of a goal in perspective of information factors.

In the farsighted model, the data attributes that are settled through recognition are addressed by the branches, while the choices about the data target are addressed in the gets out.

When “taking in” a tree, the source data is parceled into subsets in perspective of a quality regard test, which is reiterated on each one of the construed subsets recursively.

Once the subset at a center point has the proportionate motivating force as its target regard has, the recursion method will be done. We should look at an instance of various conditions that can choose on the off chance that some person should go calculating. This fuses atmosphere conditions and furthermore barometric weight conditions.

In the disentangled choice tree over, a case is grouped by dealing it with the ...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Preface
• Foreword
• Acknowledgement
• Authors
• Table of Contents
• Pre-requisite to Machine Learning
• 1. An Introduction to Machine Learning
• 2. Finding Natural Patterns in Data
• 3. Building Classification Methods
• 4. Data Pre – Processing in Python
• 5. Building Regression Models
• 6. Creating Neural Networks
• 7. Introduction to Deep Learning