Matches in SemOpenAlex for { <https://semopenalex.org/work/W3209227539> ?p ?o ?g. }
Showing items 1 to 72 of
72
with 100 items per page.
- W3209227539 abstract "A common approach to prediction and planning in partially observable domains is to use recurrent neural networks (RNNs), which ideally develop and maintain a latent memory about hidden, task-relevant factors. We hypothesize that many of these hidden factors in the physical world are constant over time, changing only sparsely. Accordingly, we propose Gated $L_0$ Regularized Dynamics (GateL0RD), a novel recurrent architecture that incorporates the inductive bias to maintain stable, sparsely changing latent states. The bias is implemented by means of a novel internal gating function and a penalty on the $L_0$ norm of latent state changes. We demonstrate that GateL0RD can compete with or outperform state-of-the-art RNNs in a variety of partially observable prediction and control tasks. GateL0RD tends to encode the underlying generative factors of the environment, ignores spurious temporal dependencies, and generalizes better, improving sampling efficiency and prediction accuracy as well as behavior in model-based planning and reinforcement learning tasks. Moreover, we show that the developing latent states can be easily interpreted, which is a step towards better explainability in RNNs." @default.
- W3209227539 created "2021-11-08" @default.
- W3209227539 creator A5001474340 @default.
- W3209227539 creator A5075919503 @default.
- W3209227539 creator A5090767927 @default.
- W3209227539 date "2021-12-06" @default.
- W3209227539 modified "2023-10-18" @default.
- W3209227539 title "Sparsely Changing Latent States for Prediction and Planning in Partially Observable Domains" @default.
- W3209227539 hasPublicationYear "2021" @default.
- W3209227539 type Work @default.
- W3209227539 sameAs 3209227539 @default.
- W3209227539 citedByCount "0" @default.
- W3209227539 crossrefType "proceedings-article" @default.
- W3209227539 hasAuthorship W3209227539A5001474340 @default.
- W3209227539 hasAuthorship W3209227539A5075919503 @default.
- W3209227539 hasAuthorship W3209227539A5090767927 @default.
- W3209227539 hasConcept C119857082 @default.
- W3209227539 hasConcept C121332964 @default.
- W3209227539 hasConcept C136197465 @default.
- W3209227539 hasConcept C147168706 @default.
- W3209227539 hasConcept C154945302 @default.
- W3209227539 hasConcept C167966045 @default.
- W3209227539 hasConcept C32848918 @default.
- W3209227539 hasConcept C39890363 @default.
- W3209227539 hasConcept C41008148 @default.
- W3209227539 hasConcept C50644808 @default.
- W3209227539 hasConcept C51167844 @default.
- W3209227539 hasConcept C62520636 @default.
- W3209227539 hasConcept C97256817 @default.
- W3209227539 hasConcept C97541855 @default.
- W3209227539 hasConceptScore W3209227539C119857082 @default.
- W3209227539 hasConceptScore W3209227539C121332964 @default.
- W3209227539 hasConceptScore W3209227539C136197465 @default.
- W3209227539 hasConceptScore W3209227539C147168706 @default.
- W3209227539 hasConceptScore W3209227539C154945302 @default.
- W3209227539 hasConceptScore W3209227539C167966045 @default.
- W3209227539 hasConceptScore W3209227539C32848918 @default.
- W3209227539 hasConceptScore W3209227539C39890363 @default.
- W3209227539 hasConceptScore W3209227539C41008148 @default.
- W3209227539 hasConceptScore W3209227539C50644808 @default.
- W3209227539 hasConceptScore W3209227539C51167844 @default.
- W3209227539 hasConceptScore W3209227539C62520636 @default.
- W3209227539 hasConceptScore W3209227539C97256817 @default.
- W3209227539 hasConceptScore W3209227539C97541855 @default.
- W3209227539 hasLocation W32092275391 @default.
- W3209227539 hasOpenAccess W3209227539 @default.
- W3209227539 hasPrimaryLocation W32092275391 @default.
- W3209227539 hasRelatedWork W1480527676 @default.
- W3209227539 hasRelatedWork W1559358587 @default.
- W3209227539 hasRelatedWork W2291973609 @default.
- W3209227539 hasRelatedWork W2799528510 @default.
- W3209227539 hasRelatedWork W2808213164 @default.
- W3209227539 hasRelatedWork W2904707853 @default.
- W3209227539 hasRelatedWork W2905438253 @default.
- W3209227539 hasRelatedWork W2916399140 @default.
- W3209227539 hasRelatedWork W2950004691 @default.
- W3209227539 hasRelatedWork W2952569483 @default.
- W3209227539 hasRelatedWork W2992832152 @default.
- W3209227539 hasRelatedWork W3012620775 @default.
- W3209227539 hasRelatedWork W3025437133 @default.
- W3209227539 hasRelatedWork W3042714747 @default.
- W3209227539 hasRelatedWork W3094736930 @default.
- W3209227539 hasRelatedWork W3105811322 @default.
- W3209227539 hasRelatedWork W3132159782 @default.
- W3209227539 hasRelatedWork W3214076633 @default.
- W3209227539 hasRelatedWork W2803743229 @default.
- W3209227539 hasRelatedWork W3148125979 @default.
- W3209227539 hasVolume "34" @default.
- W3209227539 isParatext "false" @default.
- W3209227539 isRetracted "false" @default.
- W3209227539 magId "3209227539" @default.
- W3209227539 workType "article" @default.