Basics about Machine Learning

What is Machine Learning

The ability of  computers to learn from data or there environment without being programmed explicitly 

To be more technical it's :

"A Computer program is said to learn from experience E with respect to some task T and for some Performance measure P, if it's performance on T, measure by P, imporves with experience E."                                              

~Tom Mitchell , 1997   

I know it's seems wired but that it is 

Here are some Examples :

•  The Spam Filter 

Suppose we have a program that classify between spam & not_spams on the basis of previous data that is often called classification task ( bcz we have to classify between  spams and not spams )

The above example is purely machine learning task 

Let's see why to use Machine Learning with this task 

Traditional Approach

Traditional Approach

1. First we study the problem .
2. Then write some rule like if it contains words like "for U","Credit card","free","hurry up" a lot of time may be we can observe other rules also and write a program to classify as spam on the basis of that rules .
3. Then we repeat the step 1 & 2 untill we get a good solution .
4. If it done then it's time to lunch.

Firstly the list of rules will go long and the program become so much complex to handle and also what if the spammer start using words like "4U" instead of "for u" .

Then we have to write all the rules again and again.

Therefore it's not a good idea to go for traditional method .

Now let's see Machine learning approach

Machine Learning Approach

machine Learning Approach

Automated Ml Approach

The ML approach includes following

1. Analysing the problem.
2. Identify which type of data we need. 
3. Collection data (the most important process).
4. Processing the data . 
5. Training the ML model onto the data.
6. Repeating till we get good model.
7. Deploying the model.
8. Retraining model if needed.

The process of Retraining the model is necessary because with the time the data changes thus the model need to be updated.

Fortunately this process can be automated .

thus the ML model can easily detect the changes like for U and 4U which can not be noticed in traditional approach.

There are also some other problems that have no known algorithm to solve and can be solved using Ml Approach only like Image Classification.

Machine Learning models can easily adapt to new data which is not possible in traditional method.

Continue Reading : Types of ML system