Bootstrapping Conditional GANs for Video Game Level Generation

Bootstrapping Conditional GANs for Video Game Level Generation
复制标题

自举条件 GAN 用于视频游戏关卡生成

DOI:
10.1109/cog47356.2020.9231576
复制
发表时间:
2020
期刊:
IEEE Conference on Games
影响因子:
--
通讯作者:
Togelius, Julian
Togelius, Julian
中科院分区:
--
文献类型:
--
作者:
Rodriguez Torrado, Ruben;Khalifa, Ahmed;Cerny Green, Michael;Justesen, Niels;Risi, Sebastian;Togelius, Julian

文献摘要

参考文献

被引文献

相似文献

生成对抗网络(GAN)在图像生成方面显示出了令人印象深刻的结果。然而,GAN 在生成具有某些类型约束(例如游戏关卡)的内容时面临着挑战。具体来说,很难生成既具有审美吸引力又可玩的关卡。此外,由于训练数据通常有限,因此使用当前的 GAN 生成独特的级别具有挑战性。在本文中,我们提出了一种名为条件嵌入自注意力生成对抗网络(CESAGAN)的新 GAN 架构和新的引导训练程序。 CESAGAN 是自注意力 GAN 的改进版,它结合了嵌入特征向量输入来调节判别器和生成器的训练。这允许网络对游戏对象之间的非本地依赖性进行建模,并对对象进行计数。此外,为了减少训练 GAN 所需的级别数量,我们提出了一种引导机制,将可玩的生成级别添加到训练集中。结果表明,与标准 GAN 相比,新方法不仅可以生成更多数量的可玩关卡,而且可以生成更少的重复关卡。
Generative Adversarial Networks (GANs) have shown impressive results for image generation. However, GANs face challenges in generating contents with certain types of constraints, such as game levels. Specifically, it is difficult to generate levels that have aesthetic appeal and are playable at the same time. Additionally, because training data usually is limited, it is challenging to generate unique levels with current GANs. In this paper, we propose a new GAN architecture named Conditional Embedding Self-Attention Generative Adversarial Net-work (CESAGAN) and a new bootstrapping training procedure. The CESAGAN is a modification of the self-attention GAN that incorporates an embedding feature vector input to condition the training of the discriminator and generator. This allows the network to model non-local dependency between game objects, and to count objects. Additionally, to reduce the number of levels necessary to train the GAN, we propose a bootstrapping mechanism in which playable generated levels are added to the training set. The results demonstrate that the new approach does not only generate a larger number of levels that are playable but also generates fewer duplicate levels compared to a standard GAN.
AtDELFI:自动设计清晰、完整的游戏说明
DOI: 10.1145/3235765.3235790
发表时间: 2018
期刊: Proceedings of the 13th International Conference on the Foundations of Digital Games
影响因子: --
作者:
M. Green;A. Khalifa;Gabriella A. B. Barros;Tiago Machado;Andy Nealen;J. Togelius
通讯作者: J. Togelius
走向视频游戏描述语言
DOI: 10.4230/dfu.vol6.12191.85
发表时间: 2013
影响因子: 22.7
作者:
M. Ebner;J. Levine;S. Lucas;T. Schaul;Tommy Thompson;J. Togelius
通讯作者: J. Togelius
通用视频游戏人工智能:用于评估代理、游戏和内容生成算法的多轨框架
DOI: 10.1109/tg.2019.2901021
发表时间: 2019
影响因子: 2.3
作者:
Perez-Liebana, Diego;Liu, Jialin;Khalifa, Ahmed;Gaina, Raluca D.;Togelius, Julian;Lucas, Simon M.
通讯作者: Lucas, Simon M.
通过具有进化启发式的 MCTS 使用程序角色进行自动游戏测试
DOI: 10.1109/tg.2018.2808198
发表时间: 2018
影响因子: 2.3
作者:
Christoffer Holmgård;M. Green;Antonios Liapis;J. Togelius
通讯作者: J. Togelius
一般视频游戏关卡生成
DOI: 10.1145/2908812.2908920
发表时间: 2016
期刊: Proceedings of the Genetic and Evolutionary Computation Conference 2016
影响因子: --
作者:
A. Khalifa;Diego Perez Liebana;S. Lucas;J. Togelius
通讯作者: J. Togelius