We propose Pgx, a suite of board game reinforcement learning (RL) environments written in JAX and optimized for GPU/TPU accelerators. By leveraging JAX's auto-vectorization and parallelization over accelerators, Pgx can efficiently scale to thousands of simultaneous simulations over accelerators. In our experiments on a DGX-A100 workstation, we discovered that Pgx can simulate RL environments 10-100x faster than existing implementations available in Python. Pgx includes RL environments commonly used as benchmarks in RL research, such as backgammon, chess, shogi, and Go. Additionally, Pgx offers miniature game sets and baseline models to facilitate rapid research cycles. We demonstrate the efficient training of the Gumbel AlphaZero algorithm with Pgx environments. Overall, Pgx provides high-performance environment simulators for researchers to accelerate their RL experiments. Pgx is available at https://github.com/sotetsuk/pgx.

In these days, we can take many pictures everyday and everywhere with mobile devices such as smartphones. After taking a picture, we often modify it by using some image enhancement tools so that the appearance of the picture becomes favorable to his/her own preference. However, since there are many parameters in the enhancement functions, it is not an easy task to find an appropriate parameter set to obtain the desired result. Some tools have a function that automatically determine the parameters but they do not take into account the user's preference. In this paper, we present a system to address this problem. Our system first estimates the user's preference by using RankNet. Next, the image enhancement parameters are optimized to maximize the estimated preference. We show some experimental results to demonstrate the usefulness of our system.