A new approach to training neural networks that accelerates learning

I developed and tested a new training paradigm that’s applicable to any classification problem, exploiting the class structure itself as a form of a priori information to accelerate learner training.

Professor Pascal Poupart of the Computer Science department at the University of Waterloo and I spent several months preparing this project for publication before we discovered that an engineer at Google had published a similar idea.

Article, github

Previous
Previous

Statistical analyses of trading tools for DecenTrader

Next
Next

Diversion: a 3D snake game in C++