Definition of Machine Learning

Definition:
Machine Learning is a widely used technique. Artificial Intelligence is a broad topic that consists of Machine Learning.   Machine Learning is making a machine to think like a human, work as a human, and perform better with real-life experiences.
Let us suppose that I want to create a machine (robot) which can behave like a human. For that, I need to make the machine learn different actions like talking, walking, running, etc. I will train this robot such that it will learn more actions from daily activities. So, In this example, what I have done is simply train the machine and make it learn from experiences.
So In this way, I can define Machine Learning as training machine and make it learn from experience to perform better and better.

Some Machine Learning Experts have defined the machine learning in their own ways as follow:
1. The machine is said to learn from experience with respect to some class of tasks and the performance measure P if the learner's performance at tasks in the class as measured by P improves with experience  [ Tom Michel 1997].

2.  Machine Learning is the field of study that gives computers the ability to learn without being explicitly programmed [Arthur Samuel (1959)].

Machine Learning  is used in many fields such as Medicine, Biology, Physics, Marketing, Finance,  Data mining,  Autonomous Vehicles, Handwriting Recognition, Computer vision, the Self recommendation of products, and many others.  Researches in this field Is going on. The main aim to develop different methods is to make the models easier that can be used to handle a large amount of the data. Due to the increasing amount of data, the necessity to develop this field gets enhanced.  
When we talk about Machine Learning, then the algorithms of machine learning come in the picture. There are three different algorithms of Machine Learning, namely,

1. Supervised Learning

2. Unsupervised Learning

3. Reinforcement Learning

Supervised Learning: In supervised learning, we provide input and output and try to find the relation between input and output i.e. the model that fits best with the data we are providing to the machine. We mostly use classification and regression under supervised learning. 
Suppose that I am given data of the prices of houses in certain regions with different parameters. Based on this data set, I want to predict the price of other houses taking account of all parameters used in the data set. This is a simple regression problem. First, I need to find the relation between different parameters with a price. Use this relation to predict the price of a new house.
Similarly, we can have a classification example. Suppose we want to know whether an email is a spam or not. For that, I need to check some words which frequently occur in spam emails. Based on that, we can say that email is spam or not.

Unsupervised Learning: In unsupervised learning, we input the data and we don't know the answer of the data i.e. output, the machine makes clusters/patterns on the basis of their similarities of the input data. Clustering, Association are some algorithms under unsupervised learning.
Suppose we are data set of flowers but labels are not given. In that condition, we apply unsupervised learning algorithms like clustering. Based on these clusters, we can say that a given flower belongs to that category.  

Reinforcement Learning: In reinforcement learning, we learn the control process. Let's assume that, you want to learn cycling. For that, you need to learn how to control the cycle.
The problem, due to its generality, is studied in many other disciplines, such as game theory, control theory, operations research, information theory, simulation-based optimization, multi-agent systems, swarm intelligence, statistics, and genetic algorithms.