Matches in SemOpenAlex for { <https://semopenalex.org/work/W3166763057> ?p ?o ?g. }
Showing items 1 to 86 of
86
with 100 items per page.
- W3166763057 endingPage "3744" @default.
- W3166763057 startingPage "3734" @default.
- W3166763057 abstract "Most of the recent deep reinforcement learning advances take an RL-centric perspective and focus on refinements of the training objective. We diverge from this view and show we can recover the performance of these developments not by changing the objective, but by regularising the value-function estimator. Constraining the Lipschitz constant of a single layer using spectral normalisation is sufficient to elevate the performance of a Categorical-DQN agent to that of a more elaborated rainbow{} agent on the challenging Atari domain. We conduct ablation studies to disentangle the various effects normalisation has on the learning dynamics and show that is sufficient to modulate the parameter updates to recover most of the performance of spectral normalisation. These findings hint towards the need to also focus on the neural component and its learning dynamics to tackle the peculiarities of Deep Reinforcement Learning." @default.
- W3166763057 created "2021-06-22" @default.
- W3166763057 creator A5001438167 @default.
- W3166763057 creator A5027862256 @default.
- W3166763057 creator A5039497694 @default.
- W3166763057 creator A5043910056 @default.
- W3166763057 creator A5047600807 @default.
- W3166763057 creator A5058935509 @default.
- W3166763057 date "2021-07-18" @default.
- W3166763057 modified "2023-10-09" @default.
- W3166763057 title "Spectral Normalisation for Deep Reinforcement Learning: An Optimisation Perspective" @default.
- W3166763057 hasPublicationYear "2021" @default.
- W3166763057 type Work @default.
- W3166763057 sameAs 3166763057 @default.
- W3166763057 citedByCount "4" @default.
- W3166763057 countsByYear W31667630572021 @default.
- W3166763057 crossrefType "proceedings-article" @default.
- W3166763057 hasAuthorship W3166763057A5001438167 @default.
- W3166763057 hasAuthorship W3166763057A5027862256 @default.
- W3166763057 hasAuthorship W3166763057A5039497694 @default.
- W3166763057 hasAuthorship W3166763057A5043910056 @default.
- W3166763057 hasAuthorship W3166763057A5047600807 @default.
- W3166763057 hasAuthorship W3166763057A5058935509 @default.
- W3166763057 hasConcept C108583219 @default.
- W3166763057 hasConcept C114466953 @default.
- W3166763057 hasConcept C119857082 @default.
- W3166763057 hasConcept C120665830 @default.
- W3166763057 hasConcept C121332964 @default.
- W3166763057 hasConcept C12713177 @default.
- W3166763057 hasConcept C14036430 @default.
- W3166763057 hasConcept C154945302 @default.
- W3166763057 hasConcept C168167062 @default.
- W3166763057 hasConcept C192209626 @default.
- W3166763057 hasConcept C199360897 @default.
- W3166763057 hasConcept C41008148 @default.
- W3166763057 hasConcept C5274069 @default.
- W3166763057 hasConcept C78458016 @default.
- W3166763057 hasConcept C86803240 @default.
- W3166763057 hasConcept C97355855 @default.
- W3166763057 hasConcept C97541855 @default.
- W3166763057 hasConceptScore W3166763057C108583219 @default.
- W3166763057 hasConceptScore W3166763057C114466953 @default.
- W3166763057 hasConceptScore W3166763057C119857082 @default.
- W3166763057 hasConceptScore W3166763057C120665830 @default.
- W3166763057 hasConceptScore W3166763057C121332964 @default.
- W3166763057 hasConceptScore W3166763057C12713177 @default.
- W3166763057 hasConceptScore W3166763057C14036430 @default.
- W3166763057 hasConceptScore W3166763057C154945302 @default.
- W3166763057 hasConceptScore W3166763057C168167062 @default.
- W3166763057 hasConceptScore W3166763057C192209626 @default.
- W3166763057 hasConceptScore W3166763057C199360897 @default.
- W3166763057 hasConceptScore W3166763057C41008148 @default.
- W3166763057 hasConceptScore W3166763057C5274069 @default.
- W3166763057 hasConceptScore W3166763057C78458016 @default.
- W3166763057 hasConceptScore W3166763057C86803240 @default.
- W3166763057 hasConceptScore W3166763057C97355855 @default.
- W3166763057 hasConceptScore W3166763057C97541855 @default.
- W3166763057 hasLocation W31667630571 @default.
- W3166763057 hasOpenAccess W3166763057 @default.
- W3166763057 hasPrimaryLocation W31667630571 @default.
- W3166763057 hasRelatedWork W1552148478 @default.
- W3166763057 hasRelatedWork W1553476745 @default.
- W3166763057 hasRelatedWork W2062122188 @default.
- W3166763057 hasRelatedWork W2146957157 @default.
- W3166763057 hasRelatedWork W2159600763 @default.
- W3166763057 hasRelatedWork W2159880874 @default.
- W3166763057 hasRelatedWork W2194966727 @default.
- W3166763057 hasRelatedWork W2295413729 @default.
- W3166763057 hasRelatedWork W2428834750 @default.
- W3166763057 hasRelatedWork W2626354230 @default.
- W3166763057 hasRelatedWork W2892076218 @default.
- W3166763057 hasRelatedWork W2892121731 @default.
- W3166763057 hasRelatedWork W2939075711 @default.
- W3166763057 hasRelatedWork W2950471160 @default.
- W3166763057 hasRelatedWork W2962779867 @default.
- W3166763057 hasRelatedWork W2995290757 @default.
- W3166763057 hasRelatedWork W2995638039 @default.
- W3166763057 hasRelatedWork W3032925664 @default.
- W3166763057 hasRelatedWork W3163800191 @default.
- W3166763057 hasRelatedWork W3106008061 @default.
- W3166763057 isParatext "false" @default.
- W3166763057 isRetracted "false" @default.
- W3166763057 magId "3166763057" @default.
- W3166763057 workType "article" @default.