按年被提到
20162016 · 1 次2026
强化学习与智能体›博弈与游戏 AI·arXiv:1602.02658 · cs.LG

Graying the black box: Understanding DQNs

Tom Zahavy, Nir Ben Zrihem, Shie Mannor · 2016
开始阅读 ↗PDFalphaXiv 讨论
被提及 · 2 人首次 2016-05-09 · 最近 2023-09-07
2不同的人
2活跃月份
2讨论串
181被引(OpenAlex)
arXiv 摘要(候选层不写导读)

In recent years there is a growing interest in using deep representations for reinforcement learning. In this paper, we present a methodology and tools to analyze Deep Q-networks (DQNs) in a non-blind matter. Moreover, we propose a new model, the Semi Aggregated Markov Decision Process (SAMDP), and an algorithm that learns it automatically. The SAMDP model allows us to identify spatio-temporal abstractions directly from features and may be used as a sub-goal detector in future work. Using our tools we reveal that the features learned by DQNs aggregate the state space in a hierarchical fashion, explaining its success. Moreover, we are able to understand and describe the policies learned by DQNs for three different Atari2600 games and suggest ways to interpret, debug and optimize deep neural networks in reinforcement learning.

候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
评论visarga · 2016-05

“I read a paper [1] about visualizing the Deep Q-Network used for playing Atari games. They map to 2D the game state and then colorize it by the Q scores. As a result they could visualize the strategy and observe how it is hierarchically organized in clusters containing sub-clusters. They could show regions associated with initial and termination rules. Th…”

Visualizing Large-Scale and High-Dimensional Data原评论 ↗
评论blueboo · 2023-09

“There’s no helping the incurious, but there are indeed reams of explorations unto weights and their explanations. One of my personal favourites is Zeiler and Fergus https://arxiv.org/abs/1311.2901 Tom Zahavy’s “Greying the black box” on deep RL comes to mind too https://arxiv.org/abs/1602.02658 OpenAI has their Microscope tool for investigating individ…”

Gaussian Splatting is pretty cool原评论 ↗