Sequential decision-making for the real world.
Reinforcement learning (RL) has provided solutions for several difficult sequential decision-making problems in games, locomotion, and robotics. At PFN, we aim to address challenges in reinforcement learning to enable real-world applications. Our research includes deep RL, safety, robotics, model-based RL, and imitation learning.
Reconnaissance for Reinforcement Learning with Safety Constraints
By : Shin-ichi Maeda, Hayato Watahiki, Yi Ouyang, Shintarou Okada, Masanori Koyama, Prabhat Nagarajan
Periodic Intra-Ensemble Knowledge Distillation for Reinforcement Learning
By : Zhang-Wei Hong, Prabhat Nagarajan, Guilherme Maeda
ChainerRL: A Deep Reinforcement Learning Library
By : Yasuhiro Fujita, Prabhat Nagarajan, KataokaToshiki, Takahiro Ishikawa
MANGA: Method Agnostic Neural-policy Generalization and Adaptation
By : Homanga Bharadhwaj, Shoichiro Yamaguchi, Shin-ichi Maeda