How to explain Machine Learning to a school going student?

Machine Learning and Artificial Intelligence is becoming popular day by day. Some of us are crazy about it and some are not but regardless of our interests and scope of work, it is already impacting each of our lives either directly or indirectly. From self driving cars to simple word suggestions in WhatsApp, ML/AI is being used almost everywhere. Yes, it’s not necessary for everyone to learn advance ML algorithms and other state of the art techniques but if you really want to understand how does it works in a 10 minute read then please follow this article and in next 10 minutes you will be in a position to understand the context and fundamentals of What, How and Why Machine is learning.

Before diving into the topic let us do a small activity. Let us suppose we have a problem statement.

If the price of 1Kg of sugar is 100 rupees then what would be the price of 2 Kgs of Sugar? The solution seems pretty easy but what calculation our brain did to come up with the answer and what this problem has to do with Machine Learning? Let us try to answer these questions.

In the Image below we can see that we have some input X which we are trying map to output Y, to perform this task we need some Logic L. Since the relation between X and Y is linear we can derive the Logic by calculating the ratio of quantity with price, we calculated the ratio(1:100) from our example data and applied the same logic to new data(input X) to get output Y. This is nothing but traditional Software development in which we have some input X which we parse to the Logic L and get desired output Y. If the Logic is wrong we will end up getting wrong output. Hence care should be taken while writing the Logic.

Let us modify our problem little bit to shift the topic from traditional Software development to Machine Learning. Now we have few more cases of prices w.r.t quantities.

If we follow the same procedure to calculate the output Y then the result may not be accurate because the ratio of quantity to price is now varying in our data set. If we notice closely, the relationship between X and Y is changed. Soon after the quantity is increasing beyond 5 Kg there is marginal decrease in the price. So the relationship is no more linear instead we can say it is some what quadratic. However finding a relation with less number examples is easy but what if we have millions of data?

In the third modification we will add some more features to our problem along with quantity we have quality, packaging type, etc. Now the Logic which we defined in our first problem is no longer capable of giving accurate prediction we need a more complex Logic now. There are two options to derive the Logic either sit and go through each feature and try to find the relation between input X and output Y or else let Machine learn the Logic itself by parsing the whole data set into powerful Machine Learning algorithms. Machine Learning algorithms uses mathematical functions and operations to map Input X to Output Y by learning the Logic L itself.

In the modern world, problems can be much more complicated then the one which we discussed above. For example we have different pictures of cats and dogs and our task is to classify them in two groups, here finding the relation between pictures to output Y is beyond the capabilities of a human mind. We need powerful machines to find the relation between X and Y and come up with a complex Logic. Later we can use same to logic to identify if a picture belongs to cat or a dog.

In a nutshell Machine learning is nothing but learning the Logic which maps the input problem to the output solution. Machine learning is not a new term, it was first coined around 1960’s but became popular in the last 2 decades. Since, now we have powerful machines to compute and handle huge data it is practical and convenient to let Machine find the relation and derive the Logic instead of we handcraft it. When we are dealing with millions of data then machines can outperform human by finding the relation which is not possible for a human mind to observe.

So that was all about the basic fundamental of Machine Learning, coming back to the questions we had in first paragraph. What, How and Why machine is learning. We already have answered to What and Why machine is learning. My next article will be on “How Machine is learning” and different types of Machine Learning Algorithms.