Machine learning can be described as developing a system that can extract knowledge from data to perform a specific task.
We know how to write code to complete a task. Suppose we want to write a program to tell if it is going to rain today. We may know that if the temperature is high, humidity is high, cloud cover high, it will rain. So we can design an algorithm and write an "if else clause" in Python to check the condition and tell us if it is going to rain.
Can we be sure we got the condition right and the values correct? Is it possible for us to get the computer system to look at some data about situations when there was rain and no rain to write its own algorithm for us? So we only have to give the data, and the computer can learn to do the task for us?
The traditional way of writing code is for us to write the code and ask the computer to execute it. If we go back to our problem of finding out if it is going to rain or not, we can design an algorithm, code it, and execute it. When we give input with the temperature, humidity, and cloud cover, the program will tell us if it is going to rain.
With Machine learning, we can give the data to a Machine Learning algorithm and ask the Machine Learning algorithm to develop a system or model that can find out if it is going to rain or not using the data that we have given. Once we have the model or the system, we can give input with the temperature, humidity, and cloud cover and find out if it will rain.
Machine Learning is like giving you a large collection of questions and corresponding answers and asking you to find out how you get the answer for similar questions. When you are given a question without an answer, you are able to answer the question based on what you have learned.
There are lots of applications for Machine Learning. Some of the major application areas are in Object Classification, which would help us separate Cats from Dogs or find out good or faulty products in a manufacturing plant.
We can also make predictions of temperature, stock prices, oil prices, or exchange rates using regression-type approaches.
Facebook is using ML to figure out which post out of all the posts available for you should be shown to you so that Facebook can keep you interested in continuing using Facebook without getting back to your homework.
