Matches in SemOpenAlex for { <https://semopenalex.org/work/W2890720247> ?p ?o ?g. }
- W2890720247 endingPage "9980" @default.
- W2890720247 startingPage "9971" @default.
- W2890720247 abstract "Training deep neural networks requires an exorbitant amount of computation resources, including a heterogeneous mix of GPU and CPU devices. It is critical to place operations in a neural network on these devices in an optimal way, so that the training process can complete within the shortest amount of time. The state-of-the-art uses reinforcement learning to learn placement skills by repeatedly performing Monte-Carlo experiments. However, due to its equal treatment of placement samples, we argue that there remains ample room for significant improvements. In this paper, we propose a new joint learning algorithm, called Post, that integrates cross-entropy minimization and proximal policy optimization to achieve theoretically guaranteed optimal efficiency. In order to incorporate the cross-entropy method as a sampling technique, we propose to represent placements using discrete probability distributions, which allows us to estimate an optimal probability mass by maximal likelihood estimation, a powerful tool with the best possible efficiency. We have implemented Post in the Google Cloud platform, and our extensive experiments with several popular neural network training benchmarks have demonstrated clear evidence of superior performance: with the same amount of learning time, it leads to placements that have training times up to 63.7% shorter over the state-of-the-art." @default.
- W2890720247 created "2018-09-27" @default.
- W2890720247 creator A5037989560 @default.
- W2890720247 creator A5064295764 @default.
- W2890720247 creator A5083596391 @default.
- W2890720247 date "2018-01-01" @default.
- W2890720247 modified "2023-09-27" @default.
- W2890720247 title "Post: Device Placement with Cross-Entropy Minimization and Proximal Policy Optimization" @default.
- W2890720247 cites W1502364872 @default.
- W2890720247 cites W1771410628 @default.
- W2890720247 cites W2070232376 @default.
- W2890720247 cites W2106535143 @default.
- W2890720247 cites W2123859855 @default.
- W2890720247 cites W2130942839 @default.
- W2890720247 cites W2132083787 @default.
- W2890720247 cites W2155027007 @default.
- W2890720247 cites W2163302320 @default.
- W2890720247 cites W2183341477 @default.
- W2890720247 cites W2194775991 @default.
- W2890720247 cites W2296319761 @default.
- W2890720247 cites W2525778437 @default.
- W2890720247 cites W2736601468 @default.
- W2890720247 cites W2786414509 @default.
- W2890720247 cites W2804682296 @default.
- W2890720247 cites W2952926545 @default.
- W2890720247 cites W2964308564 @default.
- W2890720247 hasPublicationYear "2018" @default.
- W2890720247 type Work @default.
- W2890720247 sameAs 2890720247 @default.
- W2890720247 citedByCount "4" @default.
- W2890720247 countsByYear W28907202472019 @default.
- W2890720247 countsByYear W28907202472020 @default.
- W2890720247 countsByYear W28907202472021 @default.
- W2890720247 crossrefType "proceedings-article" @default.
- W2890720247 hasAuthorship W2890720247A5037989560 @default.
- W2890720247 hasAuthorship W2890720247A5064295764 @default.
- W2890720247 hasAuthorship W2890720247A5083596391 @default.
- W2890720247 hasConcept C105795698 @default.
- W2890720247 hasConcept C106301342 @default.
- W2890720247 hasConcept C107321475 @default.
- W2890720247 hasConcept C11413529 @default.
- W2890720247 hasConcept C119857082 @default.
- W2890720247 hasConcept C121332964 @default.
- W2890720247 hasConcept C126255220 @default.
- W2890720247 hasConcept C137836250 @default.
- W2890720247 hasConcept C147764199 @default.
- W2890720247 hasConcept C154945302 @default.
- W2890720247 hasConcept C167981619 @default.
- W2890720247 hasConcept C19499675 @default.
- W2890720247 hasConcept C199360897 @default.
- W2890720247 hasConcept C2984842247 @default.
- W2890720247 hasConcept C33923547 @default.
- W2890720247 hasConcept C41008148 @default.
- W2890720247 hasConcept C45374587 @default.
- W2890720247 hasConcept C50644808 @default.
- W2890720247 hasConcept C52740198 @default.
- W2890720247 hasConcept C62520636 @default.
- W2890720247 hasConcept C75782508 @default.
- W2890720247 hasConcept C9679016 @default.
- W2890720247 hasConcept C97541855 @default.
- W2890720247 hasConcept C98036226 @default.
- W2890720247 hasConceptScore W2890720247C105795698 @default.
- W2890720247 hasConceptScore W2890720247C106301342 @default.
- W2890720247 hasConceptScore W2890720247C107321475 @default.
- W2890720247 hasConceptScore W2890720247C11413529 @default.
- W2890720247 hasConceptScore W2890720247C119857082 @default.
- W2890720247 hasConceptScore W2890720247C121332964 @default.
- W2890720247 hasConceptScore W2890720247C126255220 @default.
- W2890720247 hasConceptScore W2890720247C137836250 @default.
- W2890720247 hasConceptScore W2890720247C147764199 @default.
- W2890720247 hasConceptScore W2890720247C154945302 @default.
- W2890720247 hasConceptScore W2890720247C167981619 @default.
- W2890720247 hasConceptScore W2890720247C19499675 @default.
- W2890720247 hasConceptScore W2890720247C199360897 @default.
- W2890720247 hasConceptScore W2890720247C2984842247 @default.
- W2890720247 hasConceptScore W2890720247C33923547 @default.
- W2890720247 hasConceptScore W2890720247C41008148 @default.
- W2890720247 hasConceptScore W2890720247C45374587 @default.
- W2890720247 hasConceptScore W2890720247C50644808 @default.
- W2890720247 hasConceptScore W2890720247C52740198 @default.
- W2890720247 hasConceptScore W2890720247C62520636 @default.
- W2890720247 hasConceptScore W2890720247C75782508 @default.
- W2890720247 hasConceptScore W2890720247C9679016 @default.
- W2890720247 hasConceptScore W2890720247C97541855 @default.
- W2890720247 hasConceptScore W2890720247C98036226 @default.
- W2890720247 hasLocation W28907202471 @default.
- W2890720247 hasOpenAccess W2890720247 @default.
- W2890720247 hasPrimaryLocation W28907202471 @default.
- W2890720247 hasRelatedWork W2097117768 @default.
- W2890720247 hasRelatedWork W2194775991 @default.
- W2890720247 hasRelatedWork W2617019596 @default.
- W2890720247 hasRelatedWork W2618428124 @default.
- W2890720247 hasRelatedWork W2736601468 @default.
- W2890720247 hasRelatedWork W2786414509 @default.
- W2890720247 hasRelatedWork W2804682296 @default.
- W2890720247 hasRelatedWork W2943847545 @default.
- W2890720247 hasRelatedWork W2947983473 @default.
- W2890720247 hasRelatedWork W2952926545 @default.