Machine learning, a fine miracle of science, an answer to all the questions you could ask. Well, we seem to be thoroughly unsatisfied when it comes to appreciating machine learning. Nothing wrong with that; a machine that can learn all by itself and produce new things sounds pretty sweet to me. It is the go to skill to acquire if you wish to shoot your career right to the zenith. However, there is a but. The number of successful machine learning engineers has not really had any significant hike. Well, a lot more people know machine learning but they are kind of trumped when it comes to being successful. So, I thought why not try to find out what that secret ingredient is that turns your machine learning efforts towards success.
Yes, that is it, the problem. After studying the works of a number of experts, what I gathered is that it is determining and understanding the problem that sets your machine learning efforts on the path of success.
You learn to use certain algorithms; you learn to code new machine learning algorithms into life; you can use the sleekest of technologies; but, it all comes down to the identification and the proper treatment of the problem to be solved.
The big picture and the magnifying glass
If you are planning to become a machine learning expert, you are probably going to work for some business, at least initially. It will be of paramount importance for you to see the big picture of the business. That is, understanding how the market works, finding out what your employer is trying to achieve. Then you try to find out areas which might need some work.
Once you have got the big picture, it is time for you to put it under a magnifying glass. Leave your intuition behind and trust nothing but data. You narrow your vision down to find that one little corner where you can start.
Formalism and experiment
Machine learning involves exposing the program to data. We can call this data Experience and refer to it as E. The task that is to be done can be referred to as T. How well the task is done is measured in P which stands for performance. If the P increases with E while T is a constant, we consider that the machine is learning. That means if performance in the same task gets better with the increasing amount of data and passing time, we can say the machine learning effort is paying off.
The challenge is to take a small problem, identify its nitty gritties and experiment with an algorithmic model. If you can correlate the performance of your model to the performance of the company, voila, you are up for a hike.
A good time to learn
The virus outbreak has locked the nation down. Even if you are working from home, you are probably saving the time for commuting. A machine learning online course at this point might just make you thankful for the lock down.