What is your general approach when learning something new? I am not saying something relatively closer to something you already know. For instance, trying to understanding "probabilistic graphical models (PGM)", when you are not entirely confident/sure about probability in general.

Efficient: My definition here being able to understand it more intuitively, such that it can applied to a problem at hand.

Faster: Depending on the complexity, it could be anywhere between a day and a week. Of courser, certain things require more than a week. And it is more steeper to understand PGM, as it requires mastering the core concepts in probability.

My current approach:

  1. Get a gist or abstract understanding of what it does
  2. Can I understand the formula? 2.1) If yes, can I make an intuitive understanding about it? 2.2) If no, can I find a code, that implements the method/algorithm/technique.
  3. Play with the code and import it to my research.

I actively seek for a click-in-the-brain moment when trying to learn a new thing. In that way, I do not need to memorize, still, I will be able to understand it again(quicker), when I have not used that specific algorithm in a while.

Example: Linear Interpolation(lerp). For me, click-in-the-brain moment was/is, lerp parameter t is the percentage of distance between two points. A:(0,0) B:(5.0, 5.0), if t is 50% or 0.5, then you get (2.5, 2.5). Hence, you can discretize t as a time, or as a set of waypoints between A and B.

My approach works for me, if it is learning a simple equation. However, it is extremely hard to use that approach, when understanding a complex framework like PGM.

Apologies in advance, if some of the arguments are open-ended, and if they come under "depends" category. Sorry for entangling method/algorithm/technique together.

  • 1
    Similar, and possibly essentially identical, questions have been asked here before -- look at the "Related" questions at the top-right side of your posted question. That said, your "I actively seek for a click-in-the-brain moment" description might be the chunking method. More generally, possibly of interest is Encyclopedia of the Sciences of Learning edited by Norbert M. Seel (2012), which incidentally has a chapter on chunking. Commented yesterday
  • The problem with 'intuitive' and 'click-in-the-brain' is that both can be, and often are, incomplete and misleading. They often overlook details that come back to bite you unless one has really dug deep into the muck. It is like analogies - all analogies are wrong, but some are useful if you know where the minefields start. Learning is hard, there are no shortcuts.
    – Jon Custer
    Commented yesterday


Browse other questions tagged .