Machine Learning In Laymen's Terms


Machine Learning is the latest buzzword floating around. It deserves this popularity because it is one of the most interesting and most wanted sub-domains of Computer Science.
What Machine Learning really means?
By definition, Machine learning is considered a subset of Artificial Intelligence, which provides machines with the ability to learn without being explicitly programmed. Industry experts are predicting that the combination of Machine learning and hence AI and the Internet of Things (IoT) will be the new technological era setter and the businesses, startups, governments, etc. will invest huge numbers in the same.
Let’s understand Machine Learning in Laymen’s terms.
Machine Learning
I believe everyone is familiar with a famous game in Android platform named “Paper Toss”. I am addicted to this game, and I realized that the particular game is the best way to narrate “What Machine Learning is” with a Layman’s level of knowledge.
Consider you are trying to toss a paper in a bin, either in the game or real-life situation. How do you do that - Do you always shoot the paper into the bin on the first attempt itself?
I used to take multiple tries to succeed.
After the first attempt, you realize that you have to apply more force while throwing the paper. After the second attempt, you realize that the force you applied is more than required. After the two attempts, you get a clear idea of the force that needs to be applied, and in the third attempt, you find that you have to increase your throwing angle. In the fourth attempt, you will have placed the paper inside the bin.
What's happening here is that basically, after every throw, we are learning something and improving the end result. We are programmed to learn from our experience.
In Machine Learning, we do the same with machines. We can program a machine to learn from every attempt/experience/data-point and then improve the outcome. Let’s see Paper Toss example in the Machine and Non-Machine Learning approach.
Non-Machine Learning Approach
It is possible to measure the distance from the point to the bin and find out the most accurate angle to be projected. And using the values obtained, a pre-defined formula can be applied for finding out the force to be applied. Now, if you add the wind flow using a fan in the system, the system will continuously miss the target because the force and the angle of projection are pre-set using a formula that works only for a particular condition. As the conditions changed, the system has to be redesigned and reprogrammed for getting the actual result. In this particular case, the formula for force has to be upgraded by considering the direction and force of the wind along with other parameters.
Machine Learning Approach
For the same example, the ML approach begins with a generic expression for the force to be applied. But after every attempt, the system itself reframes the formula by taking the feedback. As the formula is continuously improved with taking the insights from the attempts/data points, the outcome too is improved.
Machine Learning
The most accepted technical definition of Machine Learning is
“A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.” -- Tom Mitchell, Carnegie Mellon University.
This is exactly what we observe from the paper toss experiment.