Brief

Researchers have developed a modified version of the Soft Actor-Critic (SAC) algorithm that matches the performance of Proximal Policy Optimization (PPO) in training legged robots. This new approach addresses SAC's sample inefficiency by enabling it to reuse past experiences, making it suitable for sim-to-real transfer and online learning on physical hardware. The modifications include improvements to policy initialization, critic targets, and return estimation, which allow SAC to train stably at scale across various robot platforms and locomotion tasks. AI

Researchers have developed a deep reinforcement learning framework to dynamically manage investment portfolios across global equity markets. The system, utilizing the Soft Actor-Critic algorithm, aims to optimize continuous portfolio weights by incorporating transaction costs, turnover penalties, and diversification constraints into its reward function. While the framework showed promise, particularly in the Euro Stoxx 50 and during periods of high market uncertainty, it did not consistently outperform a simple Buy and Hold strategy across all tested markets. AI