Have you ever wondered how the popular search engine ‘Google’ is able to run ahead of your finger strokes and suggest search terms for you? How amazon’s ‘Alexa’ or Apple’s ‘siri’ produces sensible responses to your voice instructions? How some of the e-mails that hit your mailbox is automatically redirected to ‘Spam’ folder?
The answer is Machine Learning (ML).
What is ML?
In ML, machines are continuously trained to learn on its own from a large amount of historical data; it then takes decisions or makes predictions based on the experience it gained from the training data rather than relying on a standard program where it is directed to perform certain actions. This is more or less similar to how a child learns from day-to-day experiences. Experience for machines means nothing but being exposed to vast quantity and variety of data. Machines learn from these experiences. For example, Google’s ambitious self-driving car project was made possible in this way. The striking difference between a child’s learning and a machine’s learning is that the machine cannot develop cognitive abilities whereas a child gains it in due course of time.
I believe I could spark your curiosity to continue reading, let me talk something about the motivation behind the upbringing of this powerful method.
A Peek into History
Making machines learn from data is not something new, in fact, the idea was introduced a long time ago. From the time computers have become an integral part of the lives of many, people wanted to make these machines intelligent and let them act wisely. The first attempt to make a machine learn was made in 1950 by Alan Turing. Since then people were able to develop algorithms to make the machines perform various tasks – making a machine decide the moves in a game of chess or performing a basic pattern recognition to name a few. By 1990, there was a paradigm shift from knowledge-driven to a data-driven approach to machine learning.
You can take a look at the below image to have an idea of the workflow in a data-driven approach.
Unfortunately, these methods did not gain popularity at that time due to lack of data. The point is, learning becomes more effective as more and more data is made available to the machine.
The relevance of ML Today
In recent times, we have been seeing an influx of data generated/transferred through media in varied forms such as written texts, images and voice records. Data is also generated by a broad range of industries covering sectors like finance, manufacturing, medical, sales, power etc. These data, if intelligently gathered and processed, can serve larger purposes like finding the degree of similarity among chunks of data or deriving information about the functioning of the system the data came from. However, as there is enormous historical data already out there, and is still continuously being generated, it is not possible for someone or a group of ‘someones’ (intends to be patented) to sit and work on these data. We need to take the help of machines to get there. You can choose a method from a long list of off-the-shelf machine learning algorithms which best suits your need.
On the whole, we are now equipped with both plenty of data and a whole lot of machine learning algorithms to work on the data.
Some of the example use cases I would like to mention here are predicting the employee churn in an organization, predicting the number of customers likely to pay the credit card bills, predicting the price sensitivity of a product, weather forecasting, classify a tumour as malignant or benign, classify a transaction as fraudulent or non-fraudulent, finding the appropriate class for a new species of plant, speech recognition, image recognition, constant monitoring of signals, finding trends in the data, detect deviations from regular trends in the data etc. etc. etc.
The Applications Of ML Are Endless………………….
Given the robustness and the wide range of applications that ML can offer, I strongly believe that it is going to bring in the remarkable change in the way industries work and the job opportunities within.