Michael Garmulewicz, a machine learning engineer part of Brain Corp’s R&D team, presented at the 2018 NeurIPS conference on December 7th, 2018.
During the conference, Garmulewicz presented a poster at the Deep Reinforcement Learning Workshop, the largest workshop at the conference. His research paper, entitled Expert-augmented actor-critic for ViZDoom and Montezuma’s Revenge, proposed an algorithm which essentially combined imitation learning with reinforcement learning to produce consistently high scoring results in Montezuma's Revenge.
The Deep Reinforcement Learning Workshop brings together researchers working at the intersection of deep learning and reinforcement learning.
An abstract for Garmulewicz’ work is available here, and the relevant poster can be found below.