Machine Learning - 101

Machine Learning is based on a big what if question. The question looks something like "What if a machine could learn like humans?". This little spark of idea where computation could be based on a model of brain cell interaction appeared in 1949 in a book titled The Organization of Behavior by Donald Hebb.

Hebb wrote, “When one cell repeatedly assists in firing another, the axon of the first cell develops synaptic knobs (or enlarges them if they already exist) in contact with the soma of the second cell.”

This small idea - acted as a very basic and essential building block for what we now call as machine learning.

Although there are multiple technical definitions of Machine Learning, I would define Machine learning as the way where we perform specific tasks without any programming.

Don't get me wrong, we still need to write programs -- but rather than writing the programs with the logical ques to solve the problem, we define a structure with which the machine could learn the logical mysteries from the previous mistakes of the problem we are trying to solve and easily scale up!

These structures are nothing but different strategies which we use to train the model - which on a high level is nothing but Math! -- more specifically Linear algebra and Calculus!

From a good set of data from which the machine could learn itself to produce the required output with improving accuracy over time. Here, the accuracy is nothing but a measure of how well the machine solves the given task. To give an analogy, this is similar to how a child learns to walk with experience 👋🏻

In traditional programming, with the understanding of the problem and the required inputs, we define the rules for the program with which the machine solves the problem and provides the output.

Whereas, in Machine Learning solution, we are trying to model the problem by giving the proper inputs and expected outputs and let the machine decide the rules within which it could operate -- and with all the experience from the data given to the machine, the machine generates a model out of it's experience which when given some inputs it produces the expected output with utmost accuracy.

That's were the concepts like, dataset, labelling, training and other come into picture -- which is a vast topic and we have lot more grounds to cover. Stay tuned for the next blog and please share your thoughts in comments!