Today I watched this lecture by Professor Patrick Winston lecture video.

I came across this series (MIT 6.034) when I was searching for a video about boosting. I have been watching the “learning” parts of the series.

Everytime I feel I may not learned much technical (e.g. may still not be able to implement the algo from sratch）but get something philosophical about learning. But I don’t worry about the techinical parts at all: there are tons of material online that could teach me how to write some algo from sratch, but there is only one Prof Winston.

For example, I learned many times how to derive “neural net” weights, I could appreciate the chain rules and functions when I saw them but I never truly get it.

From him, I suddenly learned that \(\frac{dP}{dw} = i*\delta\) which means, each weight has a partial derivative that is determined by its input and its delta, where delta depends on its ouputs.

These are valuable information that could only come from scholars like Prof Winston: it takes years of experience to provide such insightful generalization of rules/principles.

Let’s talk about today’s lecture. He talked 2 important ideas “how to be smarter” and “how to package your ideas better”. These are things we all want, right? Let’s see how to get there.

How to be smarter?

Talk to yourself when trying to solve problems.

I think this kind of explain why people are all saying “writing is important”. Perhaps part of it is writing is a way we talk to ourselves.

How to package your ideas better?

Prof Winston talked about: Your ideas are like your children, you want them to have the best life possible. I suddenly get it now: writing well and presenting well is not just a thing you do for important occasions (e.g. a job talk or a seminar with large audience), but it is also valuable for “not so important” occasions: you still want your “children” to have a good life no matter what.

This is a great lesson, it makes all the prep you do to polish and package your ideas meaningful.

As about how, you need five “S”:

• symbol (graphic) e.g. arch

• slogan (verbal) e.g. near miss

• salient (stick out) most noticeable or important.

I feel this is the hard part: how do we achieve this? But there shouldn’t be a formula for something like this.

• surprise

e.g. one shot learning is a surpise becuase people are used to algo that learn from many examples but not by one!

• story

I feel successful companies/products usually have most of these 5 components. As for me, I am not sure whether I could get somewhere like this, but these can be the standards I try to achieve even if I couldn’t get there.