We Make Machine Learning Hard
The most common questions I get about Machine Learning are: Why is it so difficult? Why it takes longer to build intuition around it?
I get it. It’s a different paradigm and needs a different approach. It requires tons of Math, Stats background. Moreover, even after knowing these, there is no straight-forward, short path to reach the optimum solution.
The development is shrouded in trial and error plus multiple POCs. We need to build many models quickly. Each model has several metrics — like precision, recall, accuracy, F1 Score and costs over epochs etc.
So not only the learning is hard but so is implementation in real life projects.
It’s hard, let’s accept that. The question is, can we stop making it even harder while learning and while implementing?
This is a TWO PART blog. In this part, we look at strategies you can use to learn Machine Learning faster. In the next post, we will dive into how to shorten the cycles of trial and error.
Can you learn Machine Learning in a jiffy?
Honestly NO! You need the conceptual background in mathematics and stats. You need to develop intuition to succeed with enterprise implementations of Machine Learning. These are slower methods and invaluable in the longer term.
But that does not mean you need years of training before you write your first code. Thanks to the tons of open source libraries and datasets available, you can learn to write a Machine Learning algorithm in a very short time.
So here are the TWO recipes — I call them The HARDER way and The LESS HARD way.
The steps in the HARDER way are needed. Mathematics and Stats are an absolute must for long term success. But should that be STEP ONE, THE BEGINNING? NO — especially so, if you want to start creating value for your organization while you develop skills.
Why is the second way less hard and faster?
Once you write code and get your hands dirty, understanding the abstract concepts and mathematics behind becomes much simpler, far more natural and fun. And you have already created something in the process. It helps build intuition faster. Most importantly, you start developing solutions faster.
So, work smart — WRITE CODE. Play with examples, check results, compare. Try harder examples. Build and learn. It’s a much more hands-on approach to learning.
After all, the fastest way to learn to play football is to PLAY FOOTBALL. Make it less hard!
Dhanashree Rajopadhye
Founder AXINOLabs — a startup specializing in AI strategy, technology consulting and up-skilling.
I am a hands-on AI researcher, strategy consultant, and a speaker at technical conferences. I conduct face to face hands-on workshops in Pune/Mumbai (India) area.