Open Internet by MindsNet

Plasticity-Stability Dilemma in Bio-plausible RL

The plasticity-stability dilemma in bio-plausible reinforcement learning (RL) is a significant challenge. Current bio-plausible methods, such as Hebbian rules, struggle with stability in non-stationary settings, leading to catastrophic forgetting. This bottleneck hinders the performance of bio-plausible agents in tasks like Pong, where adaptability and stability are crucial.

Computing & Technology, Computer Science, Machine Learning