RECORD
Object-sensitive Deep Reinforcement Learning
Machine Learning · Submitted 17 Sep 2018
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim State-of-the-art results are shown in experiments on Atari games.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Deep reinforcement learning has become popular over recent years, showing superiority on different visual-input tasks such as playing Atari games and robot navigation. Although objects are important image elements, few work considers enhancing deep reinforcement learning with object characteristics. In this paper, we propose a novel method that can incorporate object recognition processing to deep reinforcement learning models. This approach can be adapted to any existing deep reinforcement learning frameworks. State-of-the-art results are shown in experiments on Atari games. We also propose a new approach called "object saliency maps" to visually explain the actions made by deep reinforcement learning agents.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Deep reinforcement learning has become popular over recent years, showing superiority on different visual-input tasks such as playing Atari games and robot navigation. | background 1.00 | 0.05 | 0.03 | asserted 1.00 |
| s1 | Although objects are important image elements, few work considers enhancing deep reinforcement learning with object characteristics. | gap 1.00 | 0.07 | 0.03 | asserted 1.00 |
| s2 | In this paper, we propose a novel method that can incorporate object recognition processing to deep reinforcement learning models. | contribution 0.96 | 0.88 | 0.17 | asserted 0.99 |
| s3 | This approach can be adapted to any existing deep reinforcement learning frameworks. | contribution 0.61 | 0.45 | 0.07 | asserted 0.73 |
| s4 | State-of-the-art results are shown in experiments on Atari games. | result 1.00 | 0.88 | 0.48 | asserted 1.00 |
| s5 | We also propose a new approach called "object saliency maps" to visually explain the actions made by deep reinforcement learning agents. | contribution 0.98 | 0.95 | 0.09 | asserted 0.99 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Deep reinforcement learning has become popular over recent years, showing superiority on different visual-input tasks such as playing Atari games and robot navigation. Although objects are important image elements, few work considers enhancing deep reinforcement learning with object characteristics. In this paper, we propose a novel method that can incorporate object recognition processing to deep reinforcement learning models. This approach can be adapted to any existing deep reinforcement learning frameworks. State-of-the-art results are shown in experiments on Atari games. We also propose a new approach called "object saliency maps" to visually explain the actions made by deep reinforcement learning agents.
Human review
How the literature uses it
12 arXiv papers cite this in 12 sentences
Papers citing it · 2 sentences without a confident label
Cites it as background 10 papers
-
Several efforts focused on designing more interpretable model architectures and training procedures in representation learning [ref] and symbolic and relational methods [ref] , which may involve intermediate perceptual processing steps, like object recognition [ref] .
Bridging the Human-AI Knowledge Gap: Concept Discovery and Transfer in AlphaZero · Generating explanations in Reinforcement Learning · 0.91
-
Recently, several approaches involving machine learning and deep learning have also been used in the visual and language domains [ref] .
A Heterogeneous Graphical Model to Understand User-Level Sentiments in Social Media · Introduction · 0.99
-
Many works [ref] have used object-based representations as a natural way to describe the world, resulting in improved transfer and interpretability.
A Bayesian Approach to Identifying Representational Errors · Related Work · 0.91
What it cites · 2 sentences without a confident label
Uses the method or tool of 1 paper
-
We use OpenAI gym [ref] platform to perfoming our experiments.
OpenAI Gym · Experiment Settings · 1.00
Cites as background 4 papers
-
Before the introduction of object saliency, we first introduce pixel saliency [ref] .
Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps · Object Saliency Maps · 0.92
-
In the meantime, Wang et al. proposed a dueling network architecture (DuelingDQN) [ref] that decouples the state-action values into state values and action values to yield better approximation of the state value.
Dueling Network Architectures for Deep Reinforcement Learning · Deep Reinforcement Learning · 0.91
-
The burst of deep neural network over the past several years has dramatically improved object recognition capabilities such as convolutional neural network (CNN) [ref] and other variants [ref] , where the recent trend is to increase the number of layers and layer size [ref] , while using dropout [ref] to address the problem of overfitting.
Improving neural networks by preventing co-adaptation of feature detectors · Object recognition · 0.92
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.