Matches in SemOpenAlex for { <https://semopenalex.org/work/W4220652971> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W4220652971 endingPage "10" @default.
- W4220652971 startingPage "1" @default.
- W4220652971 abstract "Model-based reinforcement learning (RL) is regarded as a promising approach to tackle the challenges that hinder model-free RL. The success of model-based RL hinges critically on the quality of the predicted dynamic models. However, for many real-world tasks involving high-dimensional state spaces, current dynamics prediction models show poor performance in long-term prediction. To that end, we propose a novel two-branch neural network architecture with multi-timescale memory augmentation to handle long-term and short-term memory differently. Specifically, we follow previous works to introduce a recurrent neural network architecture to encode history observation sequences into latent space, characterizing the long-term memory of agents. Different from previous works, we view the most recent observations as the short-term memory of agents and employ them to directly reconstruct the next frame to avoid compounding error. This is achieved by introducing a self-supervised optical flow prediction structure to model the action-conditional feature transformation at pixel level. The reconstructed observation is finally augmented by the long-term memory to ensure semantic consistency. Experimental results show that our approach is able to generate visually-realistic long-term predictions in DeepMind maze navigation games, and outperforms the prevalent state-of-the-art methods in prediction accuracy by a large margin. Furthermore, we also evaluate the usefulness of our world model by using the predicted frames to drive an imagination-augmented exploration strategy to improve the model-free RL controller." @default.
- W4220652971 created "2022-04-03" @default.
- W4220652971 creator A5015912576 @default.
- W4220652971 creator A5044785404 @default.
- W4220652971 creator A5059258252 @default.
- W4220652971 creator A5078667360 @default.
- W4220652971 date "2022-01-01" @default.
- W4220652971 modified "2023-09-26" @default.
- W4220652971 title "Learning a World Model With Multitimescale Memory Augmentation" @default.
- W4220652971 doi "https://doi.org/10.1109/tnnls.2022.3151412" @default.
- W4220652971 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/35254991" @default.
- W4220652971 hasPublicationYear "2022" @default.
- W4220652971 type Work @default.
- W4220652971 citedByCount "0" @default.
- W4220652971 crossrefType "journal-article" @default.
- W4220652971 hasAuthorship W4220652971A5015912576 @default.
- W4220652971 hasAuthorship W4220652971A5044785404 @default.
- W4220652971 hasAuthorship W4220652971A5059258252 @default.
- W4220652971 hasAuthorship W4220652971A5078667360 @default.
- W4220652971 hasConcept C111919701 @default.
- W4220652971 hasConcept C119857082 @default.
- W4220652971 hasConcept C121332964 @default.
- W4220652971 hasConcept C12186640 @default.
- W4220652971 hasConcept C125411270 @default.
- W4220652971 hasConcept C126042441 @default.
- W4220652971 hasConcept C133875982 @default.
- W4220652971 hasConcept C138885662 @default.
- W4220652971 hasConcept C147168706 @default.
- W4220652971 hasConcept C154945302 @default.
- W4220652971 hasConcept C2776401178 @default.
- W4220652971 hasConcept C2776436953 @default.
- W4220652971 hasConcept C41008148 @default.
- W4220652971 hasConcept C41895202 @default.
- W4220652971 hasConcept C50644808 @default.
- W4220652971 hasConcept C61797465 @default.
- W4220652971 hasConcept C62520636 @default.
- W4220652971 hasConcept C76155785 @default.
- W4220652971 hasConcept C774472 @default.
- W4220652971 hasConcept C97541855 @default.
- W4220652971 hasConceptScore W4220652971C111919701 @default.
- W4220652971 hasConceptScore W4220652971C119857082 @default.
- W4220652971 hasConceptScore W4220652971C121332964 @default.
- W4220652971 hasConceptScore W4220652971C12186640 @default.
- W4220652971 hasConceptScore W4220652971C125411270 @default.
- W4220652971 hasConceptScore W4220652971C126042441 @default.
- W4220652971 hasConceptScore W4220652971C133875982 @default.
- W4220652971 hasConceptScore W4220652971C138885662 @default.
- W4220652971 hasConceptScore W4220652971C147168706 @default.
- W4220652971 hasConceptScore W4220652971C154945302 @default.
- W4220652971 hasConceptScore W4220652971C2776401178 @default.
- W4220652971 hasConceptScore W4220652971C2776436953 @default.
- W4220652971 hasConceptScore W4220652971C41008148 @default.
- W4220652971 hasConceptScore W4220652971C41895202 @default.
- W4220652971 hasConceptScore W4220652971C50644808 @default.
- W4220652971 hasConceptScore W4220652971C61797465 @default.
- W4220652971 hasConceptScore W4220652971C62520636 @default.
- W4220652971 hasConceptScore W4220652971C76155785 @default.
- W4220652971 hasConceptScore W4220652971C774472 @default.
- W4220652971 hasConceptScore W4220652971C97541855 @default.
- W4220652971 hasFunder F4320321001 @default.
- W4220652971 hasLocation W42206529711 @default.
- W4220652971 hasLocation W42206529712 @default.
- W4220652971 hasOpenAccess W4220652971 @default.
- W4220652971 hasPrimaryLocation W42206529711 @default.
- W4220652971 hasRelatedWork W2554067423 @default.
- W4220652971 hasRelatedWork W2585464829 @default.
- W4220652971 hasRelatedWork W2809005670 @default.
- W4220652971 hasRelatedWork W2963561234 @default.
- W4220652971 hasRelatedWork W2977487495 @default.
- W4220652971 hasRelatedWork W3022038857 @default.
- W4220652971 hasRelatedWork W3100270142 @default.
- W4220652971 hasRelatedWork W3162151154 @default.
- W4220652971 hasRelatedWork W4319083788 @default.
- W4220652971 hasRelatedWork W1629725936 @default.
- W4220652971 isParatext "false" @default.
- W4220652971 isRetracted "false" @default.
- W4220652971 workType "article" @default.