Machine Learning - more than a way to make machines learn

Machine Learning is a subset of Artificial Intelligence, as machines generally run by bits - Machine Learning is based on math and programming of it to work with 0's and 1's for making the machines work better with respect to predetermined data.

It can be simply defined as a process of fine-tuning a machine, but it shapes out better scope in finding new possibilities.

Machine Learning strikes best with respect to how we want to make the machine learn which is generally the basic rule guides called algorithm.

Algorithm - an approach towards solving or getting into a problem on a logical scale.

It can be classified into 3 ways based on Algorithm used in it.

• Supervised Learning - it can be expressed as a view of making decisions based on past data. But, technically - the outcome is a dependent variable which will always be underlying with respect to an independent variable.

For example, How to identify the phone as a smartphone - it should have a camera, touch panel, wifi, GPRS, apps, and many features. Here smartphone is a dependent variable and these features are the independent variables, helps in finding whether it is smart or not.

• Unsupervised Learning - it is not biased based on the dependencies but, it provides an idea of segmenting certain features into a category.

For example, To find the characteristics or culture of a country we can use this method to identify the activities of human that belong to the country. In another way, Amazon uses it by taking patterns out of customers buying a certain product will opt for products in the similar field by categorically finding it in buying patterns.

• Reinforcement Learning - This one is featured to an accuracy where we know the output but fine-tuning it predict the output based on the set of data.The basic concept of the trial and error method to improve the accuracy of a model.

This method is largely part of the image processing systems were a set of images will be trained and tested to predict a certain result. The famous cat identification model is built around this.

The days come along, the algorithms born every day with respect to the math - machine learning is not a new one, but with respect to data usage in the current scenario makes it more viable field to take more futuristic approach.

For further explanation on specific algorithm leave on comments