论文标题
来自Atari的像素的计划
Planning from Pixels in Atari with Learned Symbolic Representations
论文作者
论文摘要
已经证明，基于宽度的计划方法可以使用像素输入在Atari 2600域中产生最先进的性能。一种成功的方法，lolloutiw，代表了B Prolean Boolean功能集的状态。 $π$ -IW的增强版本表明，学习的功能可以与手工制作的功能具有竞争力，以进行基于宽度的搜索。在本文中，我们利用各种自动编码器（VAE）以原则性的方式直接从像素中学习功能，而无需监督。训练有素的VAE的推理模型从像素中提取布尔值，并使用这些功能推出了lutoutiw计划。最终的组合表现优于原始的Olloutiw和人类专业在Atari 2600上的表现，并大大降低了功能集的大小。
Width-based planning methods have been shown to yield state-of-the-art performance in the Atari 2600 domain using pixel input. One successful approach, RolloutIW, represents states with the B-PROST boolean feature set. An augmented version of RolloutIW, $π$-IW, shows that learned features can be competitive with handcrafted ones for width-based search. In this paper, we leverage variational autoencoders (VAEs) to learn features directly from pixels in a principled manner, and without supervision. The inference model of the trained VAEs extracts boolean features from pixels, and RolloutIW plans with these features. The resulting combination outperforms the original RolloutIW and human professional play on Atari 2600 and drastically reduces the size of the feature set.
