A little over a month ago, Simon picked up neural networks again (something he had tried a while ago but couldn’t grasp intuitively). He started the Artificial Neural Networks course on Brilliant.org and covered vectors, matrices, optimisation, perceptrons and multilayer perceptrons fairly quickly and even built his first perceptron in Python from scratch (will publish a video about this project shortly). As soon as he reached the chapter on Backpropagation, however, he realised his current knowledge of Calculus wasn’t enough. This is how Simon, completely on his own, decided to get back to studying Calculus (something he lost interest in last year). After gulping up several chapters of the Calculus Fundamentals course, Simon told me he was now ready to do Backpropagation (nearly done now). On to the convolutional neural networks (the next chapter in the course)!
As of today, these are his progress stats:
Below are some impressions of doing Calculus Fundamentals.
Sometimes, Simon loves a trick or a puzzle so much he writes it down several times over the course of a couple of days. he also really enjoys teaching this to everyone he can grab hold of.
Pythagoras Theorem Proofs
The old Gaussian trick
The kitchen drawer containing small objects like matches and tape is always a mess as Simon scavenges for treasures.
The above was inspired by Presh Talwalkar and his Mind Your Decisions. Brilliant is another source of ideas for crafty paper puzzles here:
Simon is doing an increasing load of Brilliant’s daily challenges.
Some more recent challenges:
Simon keeps thoroughly enjoying Brilliant’s approach to intelligence and learning (even though he sometimes dislikes the way the daily challenges are formulated). His latest stats:
From the courses he has done most I conclude he’s mostly into Computer Science and real world problem solving at the moment:
Below are some screen shots of the daily challenges he was especially curious about lately and also excerpts of his taking part in Brilliant’s discussions:
I noticed it’s a cyclic quadrilateral and I know that the opposite angles of a cyclic quadrilateral have to add up to 180 degrees. At first I thought: How am I even going to go about doing it, because it’s so cryptic and so full of information. But once I solved it, it actually became quite easy to draw!
The end of 2019 was packed with logic. Simon even started programming an AI that would solve logical puzzles, here is the beginning of this unfinished project (he switched to programming a chess AI instead). In the two vids below, he explains the puzzle he used as an example and outlines his plan to build the AI (the puzzles come from Brilliant.org):
And here are some impressions of Simon working on the puzzles and showing them to his sis:
Simon working on a simplified version of a search engine, including just a few documents, and performing calculations to determine how many searches one should do to make creating an index of all the documents efficient (something he has picked up in Brilliant.org’s Computer Science course.
This is an example of the learning style that Simon enjoys most. He really likes doing the daily challenges on Brilliant.org. He later sometimes discusses them with other participants or even writes wikis!
While in Southern France, Simon really enjoyed solving this puzzle (he originally saw in a Brilliant.org vid). He was so happy with his solution he kept drawing it out on paper and in digital apps, and later shared the puzzle on Twitter. This sparked quite a few reactions from fellow math lovers, encouraged Brilliant to tweet new puzzles and now Brilliant follows Simon on Twitter, how cool is that!