The Machine Learning Landscape

If you are looking to start learning about the basics of Machine learning, you are at the right place. My blog will cover overviews of various algorithms, their uses and some example code as well. Join me on this Data Science Journey, specifically around Machine Learning aspects.

There is so much material out there on the Internet on Data Science and Machine Learning.

Here is my attempt to share my knowledge and give as simplified a view as possible for all those embarking on this journey

Introduction

You would be surprised to know that this journey on Machine learning leading to Artificial Intelligence started way back in 1950. It started with the "Turing Test" originally called the "Imitation game" which was a test of a machine's ability to exhibit an intelligence that was indistinguishable from that of a human. This was followed by the Game of checkers by Arthur Samuel of IBM.

Moving on from there, the foundations of Deep Learning were set up in, as early as, 1957 through the concept of Perceptrons introduced by Frank Rosenblatt. This was followed by developments such as Nearest Neighbour's algorithm - which was the beginning of basic pattern recognition and then, the ability of a computer to pronounce like a baby in the mid-1980s.

A complete shift to data-driven learning from knowledge-driven learning happened with the IBM Deep Blue beating the World Champion at Chess, Garry Kasparov, in 1997. This breakthrough snowballed and a lot of the industry rapidly progressed since then. The modern concept of Deep Learning too was born.

The 2nd decade of the 21st century saw big leaps with all the big players in the industry contributing to the growth of ML. In 2011, IBM Watson beats two human champions in a Jeopardy competition, using a combination of Machine Learning and Natural Language Processing techniques. The other break-throughs were Google Brain being able to recognise cats on YouTube Videos, Facebook developing DeepFace that recognises faces with 97.25% accuracy and Google's AlphaGo being able to beat a Human at a very complex game called Go.

A brief timeline of the evolution of ML is shown here.

Notice that it is in recent years that it has evolved very rapidly thanks to two important break-throughs

1. Evolution of 'super computing' with commodity hardware

2. Cheap storage solutions that help in storing huge amounts of data

Machine Learning Vs AI

Both these terms are often used interchangeably. But technically there is a difference.

Machine learning, in very simple terms, is a branch of Computer Science that learns from 'Data' and can produce behaviour akin to human Intelligence. It is also a sub-domain of Artificial Intelligence (AI). So, then what is AI?

Artificial Intelligence is also a branch of Computer Science that uses various means (one of them being Machine Learning) to imitate human intelligence

ML Algorithms are categorised in various ways. Below is a mind map of the basic algorithms and related concepts that one needs to be aware of, in the ML journey.

I have shown two ways of categorising in this mind map.

1. One of them is based on the mathematical aspects of the algorithms and what is achieved by them as a result. For example Regression, Clustering etc.

2. The other way of categorising is based on 'Learning' types. Are the algorithms supervised, unsupervised etc.?

Machine Learning Basic Landscape

Here's a bird's eye view of all of the fundamental algorithms in ML. There are way more than these, I agree, but these form a good starting point

If we get fundamentals right, we could explore more in the direction of the problem we are trying to solve. Hence this is an attempt to feed into the fundamentals

You can look forward to a brief explanation of all of these in the following posts of mine.