Matches in SemOpenAlex for { <https://semopenalex.org/work/W3210531149> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W3210531149 abstract "Self-tuning algorithms that adapt the learning process online encourage more effective and robust learning. Among all the methods available, meta-gradients have emerged as a promising approach. They leverage the differentiability of the learning rule with respect to some hyper-parameters to adapt them in an online fashion. Although meta-gradients can be accumulated over multiple learning steps to avoid myopic updates, this is rarely used in practice. In this work, we demonstrate that whilst multi-step meta-gradients do provide a better learning signal in expectation, this comes at the cost of a significant increase in variance, hindering performance. In the light of this analysis, we introduce a novel method mixing multiple inner steps that enjoys a more accurate and robust meta-gradient signal, essentially trading off bias and variance in meta-gradient estimation. When applied to the Snake game, the mixing meta-gradient algorithm can cut the variance by a factor of 3 while achieving similar or higher performance." @default.
- W3210531149 created "2021-11-08" @default.
- W3210531149 creator A5007791899 @default.
- W3210531149 creator A5052796575 @default.
- W3210531149 creator A5068445186 @default.
- W3210531149 creator A5069279442 @default.
- W3210531149 creator A5075410257 @default.
- W3210531149 date "2021-11-01" @default.
- W3210531149 modified "2023-09-28" @default.
- W3210531149 title "One Step at a Time: Pros and Cons of Multi-Step Meta-Gradient Reinforcement Learning." @default.
- W3210531149 cites W1576452626 @default.
- W3210531149 cites W2002411855 @default.
- W3210531149 cites W2121863487 @default.
- W3210531149 cites W2604763608 @default.
- W3210531149 cites W2788904251 @default.
- W3210531149 cites W2916158460 @default.
- W3210531149 cites W2947999965 @default.
- W3210531149 cites W2952193948 @default.
- W3210531149 cites W2964096423 @default.
- W3210531149 cites W2979869797 @default.
- W3210531149 cites W3042532592 @default.
- W3210531149 cites W3042933240 @default.
- W3210531149 cites W3043013488 @default.
- W3210531149 cites W3104912149 @default.
- W3210531149 hasPublicationYear "2021" @default.
- W3210531149 type Work @default.
- W3210531149 sameAs 3210531149 @default.
- W3210531149 citedByCount "0" @default.
- W3210531149 crossrefType "posted-content" @default.
- W3210531149 hasAuthorship W3210531149A5007791899 @default.
- W3210531149 hasAuthorship W3210531149A5052796575 @default.
- W3210531149 hasAuthorship W3210531149A5068445186 @default.
- W3210531149 hasAuthorship W3210531149A5069279442 @default.
- W3210531149 hasAuthorship W3210531149A5075410257 @default.
- W3210531149 hasConcept C111919701 @default.
- W3210531149 hasConcept C119857082 @default.
- W3210531149 hasConcept C121955636 @default.
- W3210531149 hasConcept C134306372 @default.
- W3210531149 hasConcept C144133560 @default.
- W3210531149 hasConcept C153083717 @default.
- W3210531149 hasConcept C154945302 @default.
- W3210531149 hasConcept C162324750 @default.
- W3210531149 hasConcept C187736073 @default.
- W3210531149 hasConcept C196083921 @default.
- W3210531149 hasConcept C202615002 @default.
- W3210531149 hasConcept C2780451532 @default.
- W3210531149 hasConcept C2781002164 @default.
- W3210531149 hasConcept C33923547 @default.
- W3210531149 hasConcept C41008148 @default.
- W3210531149 hasConcept C97541855 @default.
- W3210531149 hasConcept C98045186 @default.
- W3210531149 hasConceptScore W3210531149C111919701 @default.
- W3210531149 hasConceptScore W3210531149C119857082 @default.
- W3210531149 hasConceptScore W3210531149C121955636 @default.
- W3210531149 hasConceptScore W3210531149C134306372 @default.
- W3210531149 hasConceptScore W3210531149C144133560 @default.
- W3210531149 hasConceptScore W3210531149C153083717 @default.
- W3210531149 hasConceptScore W3210531149C154945302 @default.
- W3210531149 hasConceptScore W3210531149C162324750 @default.
- W3210531149 hasConceptScore W3210531149C187736073 @default.
- W3210531149 hasConceptScore W3210531149C196083921 @default.
- W3210531149 hasConceptScore W3210531149C202615002 @default.
- W3210531149 hasConceptScore W3210531149C2780451532 @default.
- W3210531149 hasConceptScore W3210531149C2781002164 @default.
- W3210531149 hasConceptScore W3210531149C33923547 @default.
- W3210531149 hasConceptScore W3210531149C41008148 @default.
- W3210531149 hasConceptScore W3210531149C97541855 @default.
- W3210531149 hasConceptScore W3210531149C98045186 @default.
- W3210531149 hasLocation W32105311491 @default.
- W3210531149 hasOpenAccess W3210531149 @default.
- W3210531149 hasPrimaryLocation W32105311491 @default.
- W3210531149 hasRelatedWork W1848761947 @default.
- W3210531149 hasRelatedWork W2160997581 @default.
- W3210531149 hasRelatedWork W2188721481 @default.
- W3210531149 hasRelatedWork W2262925392 @default.
- W3210531149 hasRelatedWork W2405658191 @default.
- W3210531149 hasRelatedWork W2621365674 @default.
- W3210531149 hasRelatedWork W2945091653 @default.
- W3210531149 hasRelatedWork W2963973601 @default.
- W3210531149 hasRelatedWork W2974142167 @default.
- W3210531149 hasRelatedWork W2991095854 @default.
- W3210531149 hasRelatedWork W2994651240 @default.
- W3210531149 hasRelatedWork W3006333061 @default.
- W3210531149 hasRelatedWork W3008275213 @default.
- W3210531149 hasRelatedWork W3033474219 @default.
- W3210531149 hasRelatedWork W3036771316 @default.
- W3210531149 hasRelatedWork W3136455557 @default.
- W3210531149 hasRelatedWork W3157623825 @default.
- W3210531149 hasRelatedWork W3162026808 @default.
- W3210531149 hasRelatedWork W3200960336 @default.
- W3210531149 hasRelatedWork W3208204210 @default.
- W3210531149 isParatext "false" @default.
- W3210531149 isRetracted "false" @default.
- W3210531149 magId "3210531149" @default.
- W3210531149 workType "article" @default.