Disentangled (Un)Controllable Features
Not on the map. Why?
Abstract (arXiv)
In the context of MDPs with high-dimensional states, downstream tasks are predominantly applied on a compressed, low-dimensional representation of the original input space. A variety of learning objectives have therefore been used to attain useful representations. However, these representations usually lack interpretability of the different features. We present a novel approach that is able to disentangle latent features into a controllable and an uncontrollable partition. We illustrate that the resulting partitioned representations are easily interpretable on three types of environments and show that, in a distribution of procedurally generated maze environments, it is feasible to interpretably employ a planning algorithm in the isolated controllable latent partition.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 35 of its 47 references that are in the map. First posted on arXiv 31 Oct 2022.
Builds on
35 of 47 references.
References · 35
- 2014Adam: A Method for Stochastic Optimizationcited by 36,960
- 2022Generative Adversarial Networkscited by 5,603
- 2015Domain-Adversarial Training of Neural Networkscited by 3,023
- 2015Deep Reinforcement Learning with Double Q-learningcited by 1,704
- 2018Learning Latent Dynamics for Planning from Pixelscited by 1,221
- 2017Curiosity-driven Exploration by Self-supervised Predictioncited by 1,173
- 2020Mastering Atari with Discrete World Modelscited by 990
- 2018DeepMind Control Suitecited by 935
- 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
- 2020CURL: Contrastive Unsupervised Representations for Reinforcement Learningcited by 745
- 2020Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixelscited by 573
- 2020Reinforcement Learning with Augmented Datacited by 452
23 more
- 2016Reinforcement Learning with Unsupervised Auxiliary Taskscited by 400
- 2020Data-Efficient Reinforcement Learning with Self-Predictive Representationscited by 309
- 2018Towards a Definition of Disentangled Representationscited by 294
- 2019Improving Sample Efficiency in Model-Free Reinforcement Learning from Imagescited by 268
- 2021Contrastive Learning Inverts the Data Generating Processcited by 220
- 2020Never Give Up: Learning Directed Exploration Strategiescited by 210
- 2019Contrastive Learning of Structured World Modelscited by 206
- 2019DeepMDP: Learning Continuous Latent Space Models for Representation Learningcited by 204
- 2017DARLA: Improving Zero-Shot Transfer in Reinforcement Learningcited by 147
- 2020MDP Homomorphic Networks: Group Symmetries in Reinforcement Learningcited by 133
- 2017Value Prediction Networkcited by 117
- 2021A Survey on Interpretable Reinforcement Learningcited by 70
- 2022Denoised MDPs: Learning World Models Better Than the World Itselfcited by 66
- 2022Weakly Supervised Representation Learning with Sparse Perturbationscited by 64
- 2021Learning Task Informed Abstractionscited by 58
- 2020Predictive Information Accelerates Learning in RLcited by 56
- 2018Combined Reinforcement Learning via Abstract Representationscited by 37
- 2021Which Mutual-Information Representation Learning Objectives are Sufficient for Control?cited by 28
- 2021Provable RL with Exogenous Distractors via Multistep Inverse Dynamicscited by 28
- 2018Curiosity Driven Exploration of Learned Disentangled Goal Spacescited by 16
- 2017Independently Controllable Factorscited by 12
- 2021Feature-Based Interpretable Reinforcement Learning based on State-Transition Modelscited by 6
- 2021Disentanglement by Cyclic Reconstructioncited by 3
Built on by
Citing papers per year: 2026: 2.
2 papers.