Machine Learning Overview

The objective of the post is to provide a basic overview of Machine Learning and is not intended to be a technical write up.

Machine Learning is a buzz work that evokes a lot of thoughts in every ones mind . Trying to figure based on definitions provided in the Internet is not of much help in decoding the mystery of "What Machine Learning" is all about.

Machine learning in the simplest terms is the science of getting computers to act without being explicitly programmed.

It would surprise us that Machine Learning leveraged by us indirectly in our daily routine without knowing this  fact. Take the simple case of E-Mails that we use daily, the filtering of Spam Mails uses Machine Learning algorithms. Take the case of  Google News the aggregation of relevant news across the globe uses Machine Learning to identify related news items.

There are two main classifications of Machine Learning  Supervised and Unsupervised learning. An example would probably help clarify the differences between both. In Supervised learning the computer has access to training data which basically has the answers for the training data, that  is leveraged by supervised algorithms to make inferences to  other input data sets.

The Spam mail is an example of Supervisory learning where the computer is provided with training data that gives subset of values that denote spam mails. The supervised algorithms use the training data to  identify if a potential mail is a spam or not.  The algorithm also gets inputs from the  users when they it flags a normal mail  as spam and vice versa. This is called as a classification problem.

The second classification is called Unsupervised Learning. In this case there is no training data for the algorithm to leverage. An example of unsupervised learning is cluster analysis, a cluster is a group of  data that have similarities. A real world example of cluster analysis is identifying  similar customer in terms of market segments using cluster algorithms which in turn can be used for targeted marketing. In this case the clustering algorithm has no inputs to leverage and it uses the attributes in the data set to classify customer segments.

Machine Learning is leveraged in a wide variety of domains like Health Care, Pharma, Manufacturing etc.