Machine Learning Simplified : 1. Introduction
we will learn what and how's of ML. In the subsequent parts, we will explore each concept of Machine Learning in detailed way.
Photo by Markus Winkler on Unsplash
We are living in the world where data is being generated at abundant level. We are talking about more than petabytes in minutes. Now all of this big data does not provide any actionable insights aka knowledge and, we cannot also manually sit and analyze the data to get the required insights. Our goal should be getting the actionable insights with little to none human intervention.
We need powerful self learning algorithms to reach that goal. Machine Learning provides those algorithms.
Before going into definitions and jargons, let's see the applications of Machine Learning which we encounter in our daily life :
- Spam Filtering in Emails
- Price Fare prediction in Uber
- Voice agents (Alexa, Siri, Google Assistant)
- Chat bots
- Movie recommendation systems in Netflix
- Many more ...
What the heck is Machine Learning ?
The more generalized definition provided by Arthur Samuel in 1959 :
It (Machine Learning) is field of study that gives computers the ability to learn without being explicitly programmed.
It is science of programming computers that they can learn from data.
The more detailed definition is given by Tom Mitchell in 1997 :
A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.
Lets decode the above definition with spam filtering application. Here task T is filtering email into spam or ham (non spam). The experience E is email data which is already marked as spam or ham. Performance P can be count of correct predictions i.e How many emails that are predicted as spam are actually spam. From above definition, we say program is said to be learning if its performance is increasing with more data.
Why use Machine Learning ?
For some tasks, writing logic or rules in traditional programming will become tedious and could not able to solve it. For example, consider task of recognizing whether image is dog or cat which is so trivial for us, humans. But if we start writing rules suddenly we don't know where to start.
But the same task is done using Machine learning with so much ease. All we need is to provide bunch of cats and dogs images with correct labels. Ml algorithm will automatically extract the underlying patterns that will distinguish cats and dogs. Not only above example, think about how can we write a program that takes speech as input and convert it to text or vice-versa ( voice assistants).
Different types of Machine Learning
Machine Learning is broadly classified into following types based on amount of supervision during training :
- Supervised Machine Learning
In this type, the data consists of input features along with what we are trying to predict (labels). The applications either can be classification where we try to distinguish between one another or regression problems, where we need to predict the value.
Supervised learning techniques mostly used in Face Recognition, Price prediction, Spam detection, Tumors detection ...
- Unsupervised Machine Learning
This is opposite of Supervised machine learning, where inputs are not mapped with corresponding labels. Customers Segmentation, clustering of data, Anomalies detection are main applications of Unsupervised Learning.
- Semi Supervised Learning
As data becomes huge, we cannot label the entire data. Some algorithms take partially labelled data and work on them. Applications like Google photos uses these algorithms.
- Reinforcement Learning
This is one of the exciting fields in recent years. Here model is called as Agent which is in the environment and can select and perform actions, and get rewards these can be negative also (penalties). Applications like Robots, autonomous car are using these techniques
This is the end of this post. We will explore the next topics in later posts. If you want to know more about the topics, go through the hands on machine learning book. Link is added in the references section.
If you have any queries or If there are any discrepancies in the post please feel free to reach me on Twitter or linkedin