Experiences on Tutoring a Data Scientist on Reinforcement Learning
What if someone believes in you more than you deserve? The best plan would be to to add value to it. That's what happened when Shon Bhutani, a data scientist from New York believed in me to teach him Reinforcement learning on which I had no idea on.
Shon Bhutani actually contacted me to teach Bayesian networks after watching my Youtube video on Bayesian networks. I have actually written on that earlier. He is so quick in grasping concepts and more intelligent than me but yet so humble in learning. We had only two sessions on Bayesian Networks through which I taught him to work out problems so that he can code the Bayesian Networks Python library on his own. Then he wanted me to teach him LSTM but I said I am not great on that . Few weeks later, he asked if I can teach him Reinforcement learning. He sent me a soft copy of a book on reinforcement learning that he had bought . He likes reading books. He said if I can help him read through the book. I could not say "NO". Coz even I wanted to learn Reinforcement learning for a long time. The best way to learn is through teaching. I asked him a weeks time but I found it so difficult to understand. Being a slow learner, I take too much time to grasp concepts and some how I got the basics. When we started the session last week, he said he already knew the foundation (I said he is intelligent) and wanted only the mathematical part. The book he gave me just broke my head to pieces and I could not understand anything from that. I really lost hope and felt so depressed and was madly searching stuffs on Reinforcement learning for the whole last week.
When we struggle on something, especially as far as learning is concerned, you get a ray of hope when you keep struggling constantly without quitting. Of course, my slow grasping capability would pull me down sometime but somehow I manage to overcome that. And yes, you can only be suffering in learning until you meet the right teacher and Andrew NG is always one of them. My god! few days back I got his video on Markov Decision Process which lays the foundation of Reinforcement learning. That's when I stopped struggling. Listening to Andrew NG is like listening to music. You get relaxed coz you know you wont be doubtful anymore ! When u r relaxed, things get into the head in the right direction. Wow, I deciphered the whole math after watching his video. For some reason , Youtube did not get me his video for about a week. Exactly at 3 am one day, I found it :) I just admire Andrew NG a lot. Hope I teach like him one day:)
When you get the math right, you will be able to understand and solve the numerical problems. Today, I was able to teach Shon, a problem to find Optimal policy through Value iteration method. He was very convinced and happy. Those struggles, head aches and sleepless nightmares just gets vanished after delivering a satisfying session always:) Hope to learn and teach more. Let us see how it goes.
Have a great day ahead and thanks for reading.