Deep Learning

Deep Reinforcement Learning in Ice Hockey

Use the on-policy prediction method Sarsa (State–Action–Reward–State–Action) to estimate the player performance in Ice Hockey

Hybrid Training Neural-Programmer-Interpreter

A Neural-Programmer-Interpreter hybridly trained in strong-supervision mode and reinforcement-learning mode

Dual Learning Neural-Machine-Translation

Use Dual Learning on monolingual data for Neural-Machine-Translation task