论文标题
扫雷者的神经网络学习者
Neural Network Learner for Minesweeper
论文作者
论文摘要
Minesweeper是一个有趣的单人游戏，基于逻辑，内存和猜测。解决扫雷机已被证明是NP艰难的任务。确定性求解器是解决扫雷器的最著名方法。该项目提出了一个基于神经网络的学习者，用于解决扫雷器。为了选择最好的学习者，对数十万场比赛进行了培训神经网络的不同架构和配置。令人惊讶的是，拟议的基于神经网络的学习者已证明是解决扫雷器的非常好的近似功能。神经网络学习者与CSP求解器竞争良好，尤其是在游戏的初学者和中级模式下。还观察到，尽管取得了很高的成功率，但最好的神经学习者比最佳确定性求解器要慢得多。该报告还讨论了为扫雷者创建非常成功的神经网络所面临的间接费用和限制。
Minesweeper is an interesting single player game based on logic, memory and guessing. Solving Minesweeper has been shown to be an NP-hard task. Deterministic solvers are the best known approach for solving Minesweeper. This project proposes a neural network based learner for solving Minesweeper. To choose the best learner, different architectures and configurations of neural networks were trained on hundreds of thousands of games. Surprisingly, the proposed neural network based learner has shown to be a very good approximation function for solving Minesweeper. The neural network learner competes well with the CSP solvers, especially in Beginner and Intermediate modes of the game. It was also observed that despite having high success rates, the best neural learner was considerably slower than the best deterministic solver. This report also discusses the overheads and limitations faced while creating highly successful neural networks for Minesweeper.
