Matches in SemOpenAlex for { <https://semopenalex.org/work/W3138116505> ?p ?o ?g. }
- W3138116505 abstract "We propose a novel information bottleneck (IB) method named Drop-Bottleneck, which discretely drops features that are irrelevant to the target variable. Drop-Bottleneck not only enjoys a simple and tractable compression objective but also additionally provides a deterministic compressed representation of the input variable, which is useful for inference tasks that require consistent representation. Moreover, it can jointly learn a feature extractor and select features considering each feature dimension's relevance to the target task, which is unattainable by most neural network-based IB methods. We propose an exploration method based on Drop-Bottleneck for reinforcement learning tasks. In a multitude of noisy and reward sparse maze navigation tasks in VizDoom (Kempka et al., 2016) and DMLab (Beattie et al., 2016), our exploration method achieves state-of-the-art performance. As a new IB framework, we demonstrate that Drop-Bottleneck outperforms Variational Information Bottleneck (VIB) (Alemi et al., 2017) in multiple aspects including adversarial robustness and dimensionality reduction." @default.
- W3138116505 created "2021-03-29" @default.
- W3138116505 creator A5006868457 @default.
- W3138116505 creator A5010863408 @default.
- W3138116505 creator A5017009037 @default.
- W3138116505 creator A5032124306 @default.
- W3138116505 date "2021-03-23" @default.
- W3138116505 modified "2023-09-27" @default.
- W3138116505 title "Drop-Bottleneck: Learning Discrete Compressed Representation for Noise-Robust Exploration" @default.
- W3138116505 cites W1665214252 @default.
- W3138116505 cites W1686946872 @default.
- W3138116505 cites W1836465849 @default.
- W3138116505 cites W2082290707 @default.
- W3138116505 cites W2092939357 @default.
- W3138116505 cites W2117539524 @default.
- W3138116505 cites W2145339207 @default.
- W3138116505 cites W2156387975 @default.
- W3138116505 cites W2547875792 @default.
- W3138116505 cites W2548228487 @default.
- W3138116505 cites W2683470288 @default.
- W3138116505 cites W2736601468 @default.
- W3138116505 cites W2751973545 @default.
- W3138116505 cites W2788440360 @default.
- W3138116505 cites W2797548502 @default.
- W3138116505 cites W2880214242 @default.
- W3138116505 cites W2887997457 @default.
- W3138116505 cites W2913465221 @default.
- W3138116505 cites W2945774545 @default.
- W3138116505 cites W2951004968 @default.
- W3138116505 cites W2962858109 @default.
- W3138116505 cites W2963105951 @default.
- W3138116505 cites W2963857521 @default.
- W3138116505 cites W2963871073 @default.
- W3138116505 cites W2963926704 @default.
- W3138116505 cites W2964067469 @default.
- W3138116505 cites W2964083594 @default.
- W3138116505 cites W2964121744 @default.
- W3138116505 cites W2964160479 @default.
- W3138116505 cites W2964184826 @default.
- W3138116505 cites W2964185768 @default.
- W3138116505 cites W2964209830 @default.
- W3138116505 cites W2970214542 @default.
- W3138116505 cites W2996657638 @default.
- W3138116505 cites W3018703849 @default.
- W3138116505 cites W3118608800 @default.
- W3138116505 hasPublicationYear "2021" @default.
- W3138116505 type Work @default.
- W3138116505 sameAs 3138116505 @default.
- W3138116505 citedByCount "0" @default.
- W3138116505 crossrefType "posted-content" @default.
- W3138116505 hasAuthorship W3138116505A5006868457 @default.
- W3138116505 hasAuthorship W3138116505A5010863408 @default.
- W3138116505 hasAuthorship W3138116505A5017009037 @default.
- W3138116505 hasAuthorship W3138116505A5032124306 @default.
- W3138116505 hasConcept C104317684 @default.
- W3138116505 hasConcept C111030470 @default.
- W3138116505 hasConcept C11413529 @default.
- W3138116505 hasConcept C119857082 @default.
- W3138116505 hasConcept C149635348 @default.
- W3138116505 hasConcept C154945302 @default.
- W3138116505 hasConcept C185592680 @default.
- W3138116505 hasConcept C2776214188 @default.
- W3138116505 hasConcept C2780513914 @default.
- W3138116505 hasConcept C41008148 @default.
- W3138116505 hasConcept C50644808 @default.
- W3138116505 hasConcept C55493867 @default.
- W3138116505 hasConcept C59404180 @default.
- W3138116505 hasConcept C60008888 @default.
- W3138116505 hasConcept C63479239 @default.
- W3138116505 hasConcept C70518039 @default.
- W3138116505 hasConcept C73555534 @default.
- W3138116505 hasConceptScore W3138116505C104317684 @default.
- W3138116505 hasConceptScore W3138116505C111030470 @default.
- W3138116505 hasConceptScore W3138116505C11413529 @default.
- W3138116505 hasConceptScore W3138116505C119857082 @default.
- W3138116505 hasConceptScore W3138116505C149635348 @default.
- W3138116505 hasConceptScore W3138116505C154945302 @default.
- W3138116505 hasConceptScore W3138116505C185592680 @default.
- W3138116505 hasConceptScore W3138116505C2776214188 @default.
- W3138116505 hasConceptScore W3138116505C2780513914 @default.
- W3138116505 hasConceptScore W3138116505C41008148 @default.
- W3138116505 hasConceptScore W3138116505C50644808 @default.
- W3138116505 hasConceptScore W3138116505C55493867 @default.
- W3138116505 hasConceptScore W3138116505C59404180 @default.
- W3138116505 hasConceptScore W3138116505C60008888 @default.
- W3138116505 hasConceptScore W3138116505C63479239 @default.
- W3138116505 hasConceptScore W3138116505C70518039 @default.
- W3138116505 hasConceptScore W3138116505C73555534 @default.
- W3138116505 hasLocation W31381165051 @default.
- W3138116505 hasOpenAccess W3138116505 @default.
- W3138116505 hasPrimaryLocation W31381165051 @default.
- W3138116505 hasRelatedWork W2165310684 @default.
- W3138116505 hasRelatedWork W2500440664 @default.
- W3138116505 hasRelatedWork W2793035069 @default.
- W3138116505 hasRelatedWork W2885271594 @default.
- W3138116505 hasRelatedWork W2899788029 @default.
- W3138116505 hasRelatedWork W2948634622 @default.
- W3138116505 hasRelatedWork W2952884052 @default.
- W3138116505 hasRelatedWork W2971112266 @default.
- W3138116505 hasRelatedWork W2995551071 @default.