arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1805.00728cs.AIcs.NE
收藏

在深度卷积生成对抗网络的潜在空间中演化Mario水平

Evolving Mario Levels in the Latent Space of a Deep Convolutional Generative Adversarial Network

TU Dortmund University(多特蒙德工业大学);Southwestern University(西南大学);Queen Mary University of London(伦敦玛丽女王大学);University of California(加利福尼亚大学);IT University of Copenhagen(哥本哈根信息技术大学)

Vanessa Volz, Jacob Schrum, Jialin Liu, Simon M. Lucas, Adam Smith, Sebastian Risi

摘要:

生成对抗网络(GANs)是一种机器学习方法,能够在所提供的训练样本的空间上生成新颖的示例输出。用于视频游戏的关卡的程序内容生成(PCG)可以从这样的模型中受益,特别是对于存在预先存在的关卡语料库的游戏。本文训练了一个GAN,使用视频游戏关卡语料库中的关卡为超级马里奥兄弟生成关卡。该方法成功地生成了各种级别的相似的原始语料库中,但进一步提高了协方差矩阵自适应演化策略(CMA-ES)的应用。具体而言,各种适应度函数用于发现GAN的潜在空间内的最大化期望属性的级别。优化简单的静态属性,例如给定的瓦片类型分布。此外,2009年马里奥AI竞赛的冠军A* 智能体被用来评估一个关卡是否可玩,以及需要多少次跳跃动作才能击败它。这些适应度函数允许发现存在于专家设计的示例空间中的关卡,并引导搜索实现一个或多个指定目标的关卡。

英文摘要:

Generative Adversarial Networks (GANs) are a machine learning approach capable of generating novel example outputs across a space of provided training examples. Procedural Content Generation (PCG) of levels for video games could benefit from such models, especially for games where there is a pre-existing corpus of levels to emulate. This paper trains a GAN to generate levels for Super Mario Bros using a level from the Video Game Level Corpus. The approach successfully generates a variety of levels similar to one in the original corpus, but is further improved by application of the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Specifically, various fitness functions are used to discover levels within the latent space of the GAN that maximize desired properties. Simple static properties are optimized, such as a given distribution of tile types. Additionally, the champion A* agent from the 2009 Mario AI competition is used to assess whether a level is playable, and how many jumping actions are required to beat it. These fitness functions allow for the discovery of levels that exist within the space of examples designed by experts, and also guide the search towards levels that fulfill one or more specified objectives.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑