论文标题
在具体的导航任务中分析视觉表示
Analyzing Visual Representations in Embodied Navigation Tasks
论文作者
论文摘要
深度强化学习的最新进展需要大量的培训数据，并且通常会导致通常对目标任务过度专业的表示形式。在这项工作中，我们提出了一种研究这种专业化的潜在原因的方法。我们使用最近提出的投影加权规范相关分析（PWCCA）来测量通过执行不同任务在同一环境中学习的视觉表示的相似性。 然后，我们利用我们提出的方法来检查在相关但独特的导航任务上学到的视觉表示的任务依赖性。令人惊讶的是，我们发现任务的轻微差异对Squeezenet和Resnet架构的视觉表示没有可测量的影响。然后，我们从经验上证明，在一个任务上学到的视觉表示可以有效地转移到另一个任务中。
Recent advances in deep reinforcement learning require a large amount of training data and generally result in representations that are often over specialized to the target task. In this work, we present a methodology to study the underlying potential causes for this specialization. We use the recently proposed projection weighted Canonical Correlation Analysis (PWCCA) to measure the similarity of visual representations learned in the same environment by performing different tasks. We then leverage our proposed methodology to examine the task dependence of visual representations learned on related but distinct embodied navigation tasks. Surprisingly, we find that slight differences in task have no measurable effect on the visual representation for both SqueezeNet and ResNet architectures. We then empirically demonstrate that visual representations learned on one task can be effectively transferred to a different task.
