Matches in SemOpenAlex for { <https://semopenalex.org/work/W2954731415> ?p ?o ?g. }
- W2954731415 abstract "Time series forecasting is an important problem across many domains, including predictions of solar plant energy output, electricity consumption, and traffic jam situation. In this paper, we propose to tackle such forecasting problem with Transformer [1]. Although impressed by its performance in our preliminary study, we found its two major weaknesses: (1) locality-agnostics: the point-wise dot-product self-attention in canonical Transformer architecture is insensitive to local context, which can make the model prone to anomalies in time series; (2) memory bottleneck: space complexity of canonical Transformer grows quadratically with sequence length $L$, making directly modeling long time series infeasible. In order to solve these two issues, we first propose convolutional self-attention by producing queries and keys with causal convolution so that local context can be better incorporated into attention mechanism. Then, we propose LogSparse Transformer with only $O(L(log L)^{2})$ memory cost, improving forecasting accuracy for time series with fine granularity and strong long-term dependencies under constrained memory budget. Our experiments on both synthetic data and real-world datasets show that it compares favorably to the state-of-the-art." @default.
- W2954731415 created "2019-07-12" @default.
- W2954731415 creator A5034423440 @default.
- W2954731415 creator A5041666153 @default.
- W2954731415 creator A5042224802 @default.
- W2954731415 creator A5047709762 @default.
- W2954731415 creator A5048050496 @default.
- W2954731415 creator A5051342744 @default.
- W2954731415 creator A5063070498 @default.
- W2954731415 date "2019-06-29" @default.
- W2954731415 modified "2023-09-27" @default.
- W2954731415 title "Enhancing the Locality and Breaking the Memory Bottleneck of Transformer on Time Series Forecasting." @default.
- W2954731415 cites W1522301498 @default.
- W2954731415 cites W1586335931 @default.
- W2954731415 cites W1810943226 @default.
- W2954731415 cites W1815076433 @default.
- W2954731415 cites W2064675550 @default.
- W2954731415 cites W2101926813 @default.
- W2954731415 cites W2130942839 @default.
- W2954731415 cites W2172140247 @default.
- W2954731415 cites W2326533993 @default.
- W2954731415 cites W2413794162 @default.
- W2954731415 cites W2513671774 @default.
- W2954731415 cites W2519091744 @default.
- W2954731415 cites W2550848904 @default.
- W2954731415 cites W2603648311 @default.
- W2954731415 cites W2604847698 @default.
- W2954731415 cites W2607045400 @default.
- W2954731415 cites W2624190409 @default.
- W2954731415 cites W2765932895 @default.
- W2954731415 cites W2769802733 @default.
- W2954731415 cites W2773625660 @default.
- W2954731415 cites W2787214294 @default.
- W2954731415 cites W2797846142 @default.
- W2954731415 cites W2798058877 @default.
- W2954731415 cites W2798702047 @default.
- W2954731415 cites W2802023636 @default.
- W2954731415 cites W2808800115 @default.
- W2954731415 cites W2889928394 @default.
- W2954731415 cites W2891815651 @default.
- W2954731415 cites W2902854273 @default.
- W2954731415 cites W2940744433 @default.
- W2954731415 cites W2952202224 @default.
- W2954731415 cites W2963341956 @default.
- W2954731415 cites W2963403868 @default.
- W2954731415 cites W2969855422 @default.
- W2954731415 cites W3125096521 @default.
- W2954731415 hasPublicationYear "2019" @default.
- W2954731415 type Work @default.
- W2954731415 sameAs 2954731415 @default.
- W2954731415 citedByCount "6" @default.
- W2954731415 countsByYear W29547314152019 @default.
- W2954731415 countsByYear W29547314152020 @default.
- W2954731415 countsByYear W29547314152021 @default.
- W2954731415 crossrefType "posted-content" @default.
- W2954731415 hasAuthorship W2954731415A5034423440 @default.
- W2954731415 hasAuthorship W2954731415A5041666153 @default.
- W2954731415 hasAuthorship W2954731415A5042224802 @default.
- W2954731415 hasAuthorship W2954731415A5047709762 @default.
- W2954731415 hasAuthorship W2954731415A5048050496 @default.
- W2954731415 hasAuthorship W2954731415A5051342744 @default.
- W2954731415 hasAuthorship W2954731415A5063070498 @default.
- W2954731415 hasConcept C111919701 @default.
- W2954731415 hasConcept C119599485 @default.
- W2954731415 hasConcept C119857082 @default.
- W2954731415 hasConcept C127413603 @default.
- W2954731415 hasConcept C138885662 @default.
- W2954731415 hasConcept C149635348 @default.
- W2954731415 hasConcept C151406439 @default.
- W2954731415 hasConcept C165801399 @default.
- W2954731415 hasConcept C177774035 @default.
- W2954731415 hasConcept C2779808786 @default.
- W2954731415 hasConcept C2780513914 @default.
- W2954731415 hasConcept C41008148 @default.
- W2954731415 hasConcept C41895202 @default.
- W2954731415 hasConcept C66322947 @default.
- W2954731415 hasConceptScore W2954731415C111919701 @default.
- W2954731415 hasConceptScore W2954731415C119599485 @default.
- W2954731415 hasConceptScore W2954731415C119857082 @default.
- W2954731415 hasConceptScore W2954731415C127413603 @default.
- W2954731415 hasConceptScore W2954731415C138885662 @default.
- W2954731415 hasConceptScore W2954731415C149635348 @default.
- W2954731415 hasConceptScore W2954731415C151406439 @default.
- W2954731415 hasConceptScore W2954731415C165801399 @default.
- W2954731415 hasConceptScore W2954731415C177774035 @default.
- W2954731415 hasConceptScore W2954731415C2779808786 @default.
- W2954731415 hasConceptScore W2954731415C2780513914 @default.
- W2954731415 hasConceptScore W2954731415C41008148 @default.
- W2954731415 hasConceptScore W2954731415C41895202 @default.
- W2954731415 hasConceptScore W2954731415C66322947 @default.
- W2954731415 hasLocation W29547314151 @default.
- W2954731415 hasOpenAccess W2954731415 @default.
- W2954731415 hasPrimaryLocation W29547314151 @default.
- W2954731415 hasRelatedWork W153062175 @default.
- W2954731415 hasRelatedWork W2182068959 @default.
- W2954731415 hasRelatedWork W2767350163 @default.
- W2954731415 hasRelatedWork W2908168710 @default.
- W2954731415 hasRelatedWork W2964308564 @default.
- W2954731415 hasRelatedWork W2970631142 @default.
- W2954731415 hasRelatedWork W2979927445 @default.