Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287637265> ?p ?o ?g. }
Showing items 1 to 59 of
59
with 100 items per page.
- W4287637265 abstract "An artificial neural network can be trained by uniformly broadcasting a reward signal to units that implement a REINFORCE learning rule. Though this presents a biologically plausible alternative to backpropagation in training a network, the high variance associated with it renders it impractical to train deep networks. The high variance arises from the inefficient structural credit assignment since a single reward signal is used to evaluate the collective action of all units. To facilitate structural credit assignment, we propose replacing the reward signal to hidden units with the change in the $L^2$ norm of the unit's outgoing weight. As such, each hidden unit in the network is trying to maximize the norm of its outgoing weight instead of the global reward, and thus we call this learning method Weight Maximization. We prove that Weight Maximization is approximately following the gradient of rewards in expectation. In contrast to backpropagation, Weight Maximization can be used to train both continuous-valued and discrete-valued units. Moreover, Weight Maximization solves several major issues of backpropagation relating to biological plausibility. Our experiments show that a network trained with Weight Maximization can learn significantly faster than REINFORCE and slightly slower than backpropagation. Weight Maximization illustrates an example of cooperative behavior automatically arising from a population of self-interested agents in a competitive game without any central coordination." @default.
- W4287637265 created "2022-07-25" @default.
- W4287637265 creator A5021890434 @default.
- W4287637265 date "2020-10-19" @default.
- W4287637265 modified "2023-09-26" @default.
- W4287637265 title "Learning by Competition of Self-Interested Reinforcement Learning Agents" @default.
- W4287637265 doi "https://doi.org/10.48550/arxiv.2010.09770" @default.
- W4287637265 hasPublicationYear "2020" @default.
- W4287637265 type Work @default.
- W4287637265 citedByCount "0" @default.
- W4287637265 crossrefType "posted-content" @default.
- W4287637265 hasAuthorship W4287637265A5021890434 @default.
- W4287637265 hasBestOaLocation W42876372651 @default.
- W4287637265 hasConcept C119857082 @default.
- W4287637265 hasConcept C121955636 @default.
- W4287637265 hasConcept C126255220 @default.
- W4287637265 hasConcept C154945302 @default.
- W4287637265 hasConcept C155032097 @default.
- W4287637265 hasConcept C162324750 @default.
- W4287637265 hasConcept C17744445 @default.
- W4287637265 hasConcept C191795146 @default.
- W4287637265 hasConcept C196083921 @default.
- W4287637265 hasConcept C199539241 @default.
- W4287637265 hasConcept C2776330181 @default.
- W4287637265 hasConcept C33923547 @default.
- W4287637265 hasConcept C41008148 @default.
- W4287637265 hasConcept C50644808 @default.
- W4287637265 hasConcept C97541855 @default.
- W4287637265 hasConceptScore W4287637265C119857082 @default.
- W4287637265 hasConceptScore W4287637265C121955636 @default.
- W4287637265 hasConceptScore W4287637265C126255220 @default.
- W4287637265 hasConceptScore W4287637265C154945302 @default.
- W4287637265 hasConceptScore W4287637265C155032097 @default.
- W4287637265 hasConceptScore W4287637265C162324750 @default.
- W4287637265 hasConceptScore W4287637265C17744445 @default.
- W4287637265 hasConceptScore W4287637265C191795146 @default.
- W4287637265 hasConceptScore W4287637265C196083921 @default.
- W4287637265 hasConceptScore W4287637265C199539241 @default.
- W4287637265 hasConceptScore W4287637265C2776330181 @default.
- W4287637265 hasConceptScore W4287637265C33923547 @default.
- W4287637265 hasConceptScore W4287637265C41008148 @default.
- W4287637265 hasConceptScore W4287637265C50644808 @default.
- W4287637265 hasConceptScore W4287637265C97541855 @default.
- W4287637265 hasLocation W42876372651 @default.
- W4287637265 hasOpenAccess W4287637265 @default.
- W4287637265 hasPrimaryLocation W42876372651 @default.
- W4287637265 hasRelatedWork W2018374190 @default.
- W4287637265 hasRelatedWork W2048308819 @default.
- W4287637265 hasRelatedWork W2159443810 @default.
- W4287637265 hasRelatedWork W3022038857 @default.
- W4287637265 hasRelatedWork W3143722790 @default.
- W4287637265 hasRelatedWork W3180902423 @default.
- W4287637265 hasRelatedWork W3197946684 @default.
- W4287637265 hasRelatedWork W4221144246 @default.
- W4287637265 hasRelatedWork W4319083788 @default.
- W4287637265 hasRelatedWork W1629725936 @default.
- W4287637265 isParatext "false" @default.
- W4287637265 isRetracted "false" @default.
- W4287637265 workType "article" @default.