What is Machine Learning? Uses, Applications and Paradigms

Machine Learning

The ubiquitous definition of learning is the process of acquiring knowledge, developing a new skill, reflect on past experiences to unravel mistakes or recognizing people, object, figures of speech, decipher texts, conversations and so on. Pretty much how we live like, ain’t it? Now imagine how about we train a machine to do the same. Do not start thinking about robots or supercomputers or humanoids straight from the silver screens. Instead, let’s think about the web applications, mobile-based apps we use in our daily lives. Imagine your WhatsApp account itself cancelling the party on Friday because you have a presentation scheduled that very day in the late evening. Or the mathematics app solved that calculus problem you were struggling with for the past three days. You ran out of groceries. Behold, the digital assistant ordered on your behalf and you got billed too. While some of you are thinking about how convenient life would be, I bet a few of you are already wondering about this is a pure breach of privacy. Well if you think like that it is. But what if we told you, all these information transaction is going to be performed in a secure environment. Well everyone at the research laboratories are trying to solve this problem.


Well machine learning has not suddenly started to exist. For decades when you have hit the search button on the search engine, every time your pictures got tagged on Facebook, every time your keyboard recommended the next word you should type and on multiple occasions you have unknowingly used machine learning. It has existed for a quite a few decades now. So why is this sudden splurge of interest in Machine Learning among the masses now? If you think carefully, machine learning algorithms take a lot of time to train because it trains with a lot of data and hence needs sophisticated hardware. Nowadays, computing devices are available at cheaper rates and our digital footprint has grown from 130 billion gigabytes in 1995 to approximately 40 trillion gigabytes today. Hence people have a reinvigorated interest to implement machine learning in all walks of life to make the mundane jobs simpler and in order synchronizing their lives.

Technically machine learning is the process of establishing relationships between the attributes in a dataset. Machine learning accounts for learning from the data and more importantly adapting itself according to the data. It saves programmers from writing down rules to automate certain tasks which leads to the improvement of overall efficiency of the system because anomalies can be dealt with easily. There are three broad paradigms in machine learning. Supervised Learning, Unsupervised Learning and Reinforcement Learning. In Supervised Learning, the training data is already labelled and the algorithm works in a way to match these labels. In Unsupervised Learning, the primary purpose is to group identical elements to specific classes. Reinforcement learning is learning by trial and error. Here the (learning) agents themselves interact with the environment and learn by getting positive or negative rewards. It is a more natural way of learning and the technique behind autonomous cars, self-flying helicopters and the very famous AlphaGo which beat the world champion Lee Sedol in the game of Go. What is similar between all of the three methods is that they all seek to predict future events based on the data they were trained on.

A host of several learning methods are available which works on different scenarios. You cannot apply the same method to solve all kinds of problems. It takes real skilled professionals to identify which method will work on the given scenario. To know more about these methods, join our course “Essentials of Machine Learning using Python” where you are not only taught about the Fundamentals of Machine Learning, but also how you program them to suit your project’s needs.