Machine Learning — in easy and simple words .

Machine learning is an ability of a computer to learn without being particularly programmed . We need not give particular instructions , we give the data and instead of a predefined output we provide the computer with an machine learning algorithm. The programmer inputs data , chooses an algorithm and tells the computer about errors and makes corrections , but its the machine that makes the decision. It finds patterns , makes decisions and gains insights .

We have a problem ,

• We need to create a rule that might solve the problem .
• then apply the rule.
• then take feedback about whether the solution was correct or not .
• If correct add it to memory or if incorrect adjust the rule and then check and add it to memory.

Machine learning is categorised in 4 categories :

1. Supervised learning — This is applied when you know enough about your data . You can help the machine to understand what is the data that it has . In this kind of learning you give labelled data(tagged with identifying information)to the machine and pre-determined correct output. For example — A tutor teaches you how to solve the problem , solves some questions in front of you, shows you the method and then asks you to solve another question .Over time machine will learn and adapt its model to improve its output.
2. Un-supervised learning — This is applied when you don’t know much about the data .Input data is not labelled and no pre-determined output is given . You use different algorithms to let machine create connections by studying and observing the data . For example — You are new to a class , you see other students solving problems , you try to observe how your classmates are solving and try to learn . After observing for quite some time you try to connect the dots and are able to solve the problem.
3. Semi-supervised learning — It is a combination of supervised and un-supervised . Initially a small set of data is fed to the machine that is labelled (as in supervised learning) , allow the machine to study and observe the rest of the data by finding patterns and gaining insights .(as in un-supervised learning). Semi-supervised learning is not that common but it works well when data is so large that its not practical to use supervised learning . This method is used when other two learnings are difficult to use .
4. Reinforcement — In above three learnings , we are trying to find the best model that can accurately classify datasets and find meaningful groups. But in reinforcement , instead of letting machine to observe and learn we are giving the machine a clear goal. Rewards for hitting the goals and other rewards for not hitting the goal. It finds best strategies to get more rewards . Reinforcement learning allows machine to grow beyond our understanding . This helps in skipping steps that are required in unsupervised learning .

This is brief description about machine learning to give you an idea how machine learns and and in what ways .