Jinwoo - Machine Learning For Beginners: The Simplified Guide to Understanding Machine Learning

INTRODUCTION

Thank you for downloading the book Machine learning for beginners The purpose of this book is to teach you about how machine learning works easily, using step by step approach and without mathematics or computer language-- Deep Neural Networks

This book is about guiding the complete beginners how Machine Learning works using perceptron which is similar to neurons in the human brain with the basic knowledge to know in order to study Machine Learning easily.

ML is a tricky subject but this book seeks to introduce the subject matter to complete beginners such that the basics and theory of ML as to its operation and uses can easily be comprehended by complete beginners. This book will break to small bits the essentials of ML to facilitate the easy understanding of Machine Learning.

The benefits of Machine Learning

Here are some benefits of Machine Learning. We shall deal more exhaustively with these benefits under another chapter on the application of Machine Learning

Machine Learning helps in managing massive and multidimensional data of different types in a particular environment. It makes computer handling of assorted information from different origin to be easy such that they can be stockpiled and arranged in such a manner to give fast outcomes for different searches.

Machine Learning makes information processing very fast, diverse, multidimensional, accurate and cheap while dissecting on a large scale mind-blowing and massive information.

By gathering information and feeding it into the computer, Machine Learning can be used to forecast and to make the high precision prediction on available opportunities as well as warnings that can lead to abstaining from potential dangers in the business world. This kind of information gathering and computing can give a business a competitive edge over a stiff contender.

It helps in the financial world with critical decisions on two core areas which are, when to venture into a business or forestall extortion which may help to make the decision on when to exchange or abstain from the business.

It can help in limiting of data fraud in security outfits of government or privately owned ones. Hence Machine Learning is of top use in security systems.

Machine Learning can be of tremendous use in healthcare by reason of data input. The machine can be used to make a precise and accurate diagnosis.

It is becoming the future of retail by virtue of your past buying the computer can elucidate the product the individual may likely be interested in. That is the new port of the retail world. Machine Learning can now process the history of former purchases and predict what the client is likely going to be interested in.

It has a huge usage in oil and gas where Machine Learning is used to process available data into intelligent information streamlining oil dissemination to make it more productive and precise. The usage is still expanding in this area.

Its of huge importance in transportation businesses. It helps to distinguish between patterns which are of great benefit in transportation. Efficient use of Machine Learning can help to expand benefits.

CHAPTER 1

WHAT IS MACHINE LEARNING?

Machine Learning is an arm of artificial intelligence which is a relatively new form of computer programming that allows the computer to access massive data which also allows the machine to learn on its own through the experience without the machine being programmed as such. The machine learns on its own. Through different input data and through experience it is able to do certain tasks which were not pre-programmed. It is basically the analysis and interpretation of patterns of information supplied to the computer.

Machine Learning otherwise known as ML is a growing subject which can play key roles in a wide range of conditions from language to lineage recognition to data mining.

The mainstay here is automatic learning of computers due to experience on data provided. It puts the computer in a self-learning mode as the computer is fed with new data, it readjusts, grows and develop itself irrespective of the fact that it was not programmed explicitly. It gains from examples and information fed it and comes to a resolution with little no human aid

THE TYPICAL TASKS ENTRUSTED TO MACHINE LEARNING

Machine Learning tasks can be classified into many broad categories. Under Machine Learning tasks we have supervised learning and semi-supervised learning. Supervised learning is built around mathematical models which can be derived from a set of data input and desired output. So we have under this category of training data.

Semi-supervised learning is the mathematical model built around incomplete data training. That means some of the input data do not have labels attached to them.

There are many types of tasks that are associated with Machine Learning but the key ones are:

Feature selection

Regression

Classification

Clustering

Testing and matching

Density estimation

Dimension reduction

Multivariate querying

Of all these tasks associated with Machine Learning, we will like to concentrate on regression and classification. Regression and classification tasks are basically supervised learning. While output for classification is discrete in nature regression is a prediction of continuous quantities.

REGRESSION TASK

This task under Machine Learning has to do with numerical estimation and data that are continuous in nature which can otherwise be known as continuous variables. Under this task, we have things like

Estimation price for a housing unit

Product price

Stock price etc

This task has to do with the financial world of buying and selling and accounting procedures associated with numerical data.

The following methods can be used to achieve the objectives of this task which can be subdivided into two groups.

High accuracy method:

• Kernel regression
• Gaussian process regression

Not so high accurate method

• Regression trees
• Linear regression
• Support vector regression
• LASSO.

The other branch of Machine Learning we will like to treat is

CLASSIFICATION TASK

This is otherwise known as discrete variables. It is about predicting a category of data. Example of this is predicting whether a mail is a spam or not. It is also used highly in the area of healthcare. That is, predicting whether or not a person is suffering from one disease or not. It is also used in the financial world to determine whether a transaction is fraudulent or not.

This task can also be divided into two broad branches namely the high accuracy and the not so accurate.

Under high accuracy tasks, the following methods could be used to solve the problems of this task

• K Nearest Neighbours
• Artificial neural networks (ANN)
• Support vector machine (SVM)
• Random forest

Not so accurate method

• Decision trees
• Boosted trees
• Logistic regression
• Naive Bayes
• Deep learning.

Under the topic Regression we will like to expatiate on the following:

KERNEL REGRESSION

This is an estimation technique to fit data. It is categorized as a non parametric technique. It is different from linear regression which is based on the assumption of normal distribution because it does not assume any distribution to estimate the regression function.

What kernel regression does is to put a set of the identical function called kernel local to each observation data point. It is a superset of other weighted regression and it is closely related to Moving Average (MA), K nearest Neighbour (KNN), Radial Basis Function (RBF), neural network (NN) and Support Vector Machine (SVM)