A Multi-Decomposition Method for Compressing Larger AI Models Based on Reinforcement Learning

This paper proposes an innovative model compression scheme that combines various decomposition methods with a reinforcement learning-based optimization framework, jointly optimizing model compression rate and accuracy.