Evolutionary Optimization of Deep Learning Agents for Sparrow Mahjong

TL;DR

This paper introduces Evo-Sparrow, a deep learning agent for Sparrow Mahjong trained via evolutionary strategies, outperforming rule-based agents and matching PPO performance, showcasing a hybrid approach for complex game AI.

Contribution

It presents a novel hybrid method combining deep learning and evolutionary optimization for decision-making in stochastic, partially observable games.

Findings

Outperforms rule-based agents in Sparrow Mahjong

Achieves performance comparable to PPO baseline

Demonstrates effectiveness of evolutionary strategies in complex games

Abstract

We present Evo-Sparrow, a deep learning-based agent for AI decision-making in Sparrow Mahjong, trained by optimizing Long Short-Term Memory (LSTM) networks using Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Our model evaluates board states and optimizes decision policies in a non-deterministic, partially observable game environment. Empirical analysis conducted over a significant number of simulations demonstrates that our model outperforms both random and rule-based agents, and achieves performance comparable to a Proximal Policy Optimization (PPO) baseline, indicating strong strategic play and robust policy quality. By combining deep learning with evolutionary optimization, our approach provides a computationally effective alternative to traditional reinforcement learning and gradient-based optimization methods. This research contributes to the broader field of AI game playing, demonstrating the viability of hybrid learning strategies for complex stochastic games. These findings also offer potential applications in adaptive decision-making and strategic AI development beyond Sparrow Mahjong.