DAE neural net and Keras

Simon told me today he was ready to start building his own DAE (Denoising Auto Encoder) neural network. He said he would be using a documentation page about a machine learning library called Keras at blog.keras.io/building-autoencoders-in-keras.html. He found this documentation page completely on his own, by searching the web and digging into Python fora. I just watch him google something like “How can I install Keras 1.0?” and find GitHub discussions on the subject that guide him along. Or I see him type “How to install Python on Windows?” and follow the instructions at How-to-Geek. Eventually, he came up with a list of steps that he needed to complete in order to be able to install Keras (installing Python 3 (looking up why it should be 3 and not 2), installing PIP, installing Tensor Flow, etc). It didn’t work on a mac laptop, so he tried everything on a pc and it worked! Two steps required that he used terminal. It was amazing to see him plan ahead, search and implement (notoriously difficult) steps completely independently.

He has started working on the DAE tonight. Working on a node package that makes `manifest.json` files (for Chrome extensions) at the same time, so not sure he’ll finish soon. “Mom, I’ve got so many things to do! There are so many thoughts in my head!”



Problem with a Genetic Algorithm Game

Simon has been working very hard at making his version of the Citius Invaders game that Siraj Raval presented in Python during Week 9 of his “Math of Intelligence” course. It’s a simplified version that still incorporates a genetic algorithm. Simon’s plan is to code the same game in P5 during his next live tutorial (at 17:00 CET on 30 November). Unfortunately, after many lines of code in several files, when the game was nearly ready, he faced a problem he doesn’t know how to resolve. He tried getting some online help via the Coding Train Slack channel but didn’t get much feedback. The code is now on github at: https://github.com/simon-tiger/citius-invaders/

Here is how Simon describes the problem:

Hi! I’m working on a “Citius Invaders” game, and I have a problem.
In the game, there are enemies, a spaceship, and bullets that it shoots.
The parameters of the game are:
* The game starts with 6 enemies.
* The minimal amount of enemies is 4.
* The maximal amount of enemies is 100.
The rules of the game are:
* When a bullet collides with an enemy, the bullet and the enemy get deleted.
* If there are 4 enemies on the screen, you’re not allowed to shoot any bullets any more.
* If there are 100 enemies on the screen, you’ve lost the game.
* The enemies reproduce every 5 seconds (this amount of time decreases every generation).
The goal of the game is to stay as close to 4 enemies as possible.
The problem is that: After the 5 seconds, the enemies don’t stop crossing over! I think the order that I’m crossing them over is wrong. They breed (cross over) forever and that makes the game pause forever. Maybe the algorithm in which I let them cross over is wrong. I have a `while` loop somewhere, maybe it became a forever loop. The `while` loop is at line 52 in `Population.pde`:

 while (!selectedEnemies1.isEmpty() || !selectedEnemies2.isEmpty()) {
 int index1 = int(random(selectedEnemies1.size()));
 int index2 = int(random(selectedEnemies2.size()));
 if (index1 == index2) {
 index2 = (index2 + 1) % selectedEnemies2.size();

Below are a some videos showing Simon at different stages of the project. In the top video, he speaks of the difficulties he encountered with the frame rate and the genetic algorithm.



Simon working on a neural networks paper

Simon was working on a neural networks paper in Jupyter Notebook on Friday evening, but didn’t finish it because the Coding Train live stream started. He says he can no longer continue without having to do too much copy-pasting from this version into a new one, as his in-your-browser time expired, so I’m posting some screen shots of the unfinished paper below. This is the way Simon teaches himself: he follows lectures and tutorials online and then goes ahead to writing his own “textbook”or recording his own “lecture”. Much of the knowledge he acquires on neural networks these days comes from Siraj Raval’s YouTube series “The Math of Intelligence”.


Neural Networks Paper Jupyter 2017-11-20 1

Neural Networks Paper Jupyter 2017-11-20 2

Neural Networks Paper Jupyter 2017-11-20 3

Neural Networks Paper Jupyter 2017-11-20 4

Simon building a Perceptron in Processing

Simon has already built a Perceptron before, several months ago, while following along with Daniel Sgiffman’s Coding Train channel. This time around, he is writing his own code ad doing all the matrix calculations himself. He hasn’t finished programming this network yet, but it’s a good start:



Doing Matrices in Khan Academy’s Precalculus course:

Schermafbeelding 2017-11-16 om 12.34.36

Simon’s bedtime lectures on neural networks



There’s a part 3 coming!



“Mom, my ClickCharts trial period expired, so I found this Virtual Paradigm Enterprise!” (Simon independently searches for free options to make beautiful diagrams online).

Here a diagram of an LSTM neural network:

LSTM Cell in Virtual Paradigm Enterprise 19 Nov 2017 2

And an RNN:

RNN Cell in Virtual Paradigm Enterprise 19 Nov 2017

Simon’s Live Streams (Speechjs and Supershape Morphing)

Simon’s second Live Stream today (in two parts, see the archived versions below) was a big success! He fixed the bug in his own Speechjs library (for speech recognition and speech synthesis), demonstrated it in action and (in the second part) created some awesome looking supershapes by morphing one supershape into another.

Please, vote for Simon’s speech recognition library on Strawpoll: https://strawpoll.com/e55esk3h

Just another day in graphs

Simon loves looking at things geometrically. Even when solving word problems, he tends to see them as a graph. And naturally, since he started doing more math related to machine learning, graphs have occupied an even larger portion of his brain! Below are his notes in Microsoft Paint today (from memory):

Slope of Line:

Slope of Line 15 November 2017

Steepness of Curve:

Steepness of Curve 15 November 2017

An awesome calculator Simon discovered online at desmos.com/calculator that allows you to make mobile and static graphs:

Desmos.com Polynomial 15 Nov 2017

Desmos.com Polynomial 15 Nov 2017 1

Yesterday’s notes on the chi function (something he learned through 3Blue1Brown‘s videos on Taylor polynomials):


Simon following The Math of Intelligence course by Siraj Raval:



Sphere Morphing in Processing

Simon is trying to write a program for Sphere Morphing in Processing, first making a test code in p5.js (available here: https://alpha.editor.p5js.org/simontiger/sketches/S1zcwevkz)

In the video below, Simon is explaining the challenge using Magformers triangles:


Unfortunately, the test code doesn’t quite work yet: Simon is getting three infinite triangles around the circle.