Justin DuJardin

Justin DuJardin

I like open source, games, building tools, and solving problems with machine learning.

Research#

Open source#

Math problems solved step by step, with the work shown. A computer algebra system supplies the rules; reinforcement learning picks the next one.

ng2-material

deprecated

Components on top of Angular Material, from the Angular 2 beta days.

2015 to 2018626 stars

In other people’s projects:

Open questions#

  • Can a neural network learn to add, subtract, multiply, and divide exactly, and keep doing so far outside the numbers it trained on? Stable Because Stuck
  • When a network learns a representation, what is actually in it, and how much of that could be predicted before training or edited after?

If you fund or collaborate on this kind of work, write to me.