Minigo: An Exercise in Reproducibility for Machine Learning

AlphaGoZero is an incredible accomplishment in machine learning - using pure reinforcement learning to learn to play Go from scratch. Minigo is a project that reproduces the AlphaGoZero paper.

In this talk, Brian Lee will discuss the difficulties encountered in reimplementing AlphaGoZero, and the lessons to be learned in machine learning research, publication, and productionization.

Schedule of events
12:00 - 12:15 pm - Set up and introductions
12:15 - 1:00 pm - Featured Talk
1:00 - 1:45 pm - Questions and Networking