Martin Falk, Adam Strupp, Benjamin Scellier, Arvind Murugan
{"title":"通过非平衡记忆进行对比学习","authors":"Martin Falk, Adam Strupp, Benjamin Scellier, Arvind Murugan","doi":"arxiv-2312.17723","DOIUrl":null,"url":null,"abstract":"Learning algorithms based on backpropagation have enabled transformative\ntechnological advances but alternatives based on local energy-based rules offer\nbenefits in terms of biological plausibility and decentralized training. A\nbroad class of such local learning rules involve \\textit{contrasting} a clamped\nconfiguration with the free, spontaneous behavior of the system. However,\ncomparisons of clamped and free configurations require explicit memory or\nswitching between Hebbian and anti-Hebbian modes. Here, we show how a simple\nform of implicit non-equilibrium memory in the update dynamics of each\n``synapse'' of a network naturally allows for contrastive learning. During\ntraining, free and clamped behaviors are shown in sequence over time using a\nsawtooth-like temporal protocol that breaks the symmetry between those two\nbehaviors when combined with non-equilibrium update dynamics at each synapse.\nWe show that the needed dynamics is implicit in integral feedback control,\nbroadening the range of physical and biological systems naturally capable of\ncontrastive learning. Finally, we show that non-equilibrium dissipation\nimproves learning quality and determine the Landauer energy cost of contrastive\nlearning through physical dynamics.","PeriodicalId":501325,"journal":{"name":"arXiv - QuanBio - Molecular Networks","volume":"25 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2023-12-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Contrastive learning through non-equilibrium memory\",\"authors\":\"Martin Falk, Adam Strupp, Benjamin Scellier, Arvind Murugan\",\"doi\":\"arxiv-2312.17723\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Learning algorithms based on backpropagation have enabled transformative\\ntechnological advances but alternatives based on local energy-based rules offer\\nbenefits in terms of biological plausibility and decentralized training. A\\nbroad class of such local learning rules involve \\\\textit{contrasting} a clamped\\nconfiguration with the free, spontaneous behavior of the system. However,\\ncomparisons of clamped and free configurations require explicit memory or\\nswitching between Hebbian and anti-Hebbian modes. Here, we show how a simple\\nform of implicit non-equilibrium memory in the update dynamics of each\\n``synapse'' of a network naturally allows for contrastive learning. During\\ntraining, free and clamped behaviors are shown in sequence over time using a\\nsawtooth-like temporal protocol that breaks the symmetry between those two\\nbehaviors when combined with non-equilibrium update dynamics at each synapse.\\nWe show that the needed dynamics is implicit in integral feedback control,\\nbroadening the range of physical and biological systems naturally capable of\\ncontrastive learning. Finally, we show that non-equilibrium dissipation\\nimproves learning quality and determine the Landauer energy cost of contrastive\\nlearning through physical dynamics.\",\"PeriodicalId\":501325,\"journal\":{\"name\":\"arXiv - QuanBio - Molecular Networks\",\"volume\":\"25 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-12-29\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - QuanBio - Molecular Networks\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2312.17723\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - QuanBio - Molecular Networks","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2312.17723","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Contrastive learning through non-equilibrium memory
Learning algorithms based on backpropagation have enabled transformative
technological advances but alternatives based on local energy-based rules offer
benefits in terms of biological plausibility and decentralized training. A
broad class of such local learning rules involve \textit{contrasting} a clamped
configuration with the free, spontaneous behavior of the system. However,
comparisons of clamped and free configurations require explicit memory or
switching between Hebbian and anti-Hebbian modes. Here, we show how a simple
form of implicit non-equilibrium memory in the update dynamics of each
``synapse'' of a network naturally allows for contrastive learning. During
training, free and clamped behaviors are shown in sequence over time using a
sawtooth-like temporal protocol that breaks the symmetry between those two
behaviors when combined with non-equilibrium update dynamics at each synapse.
We show that the needed dynamics is implicit in integral feedback control,
broadening the range of physical and biological systems naturally capable of
contrastive learning. Finally, we show that non-equilibrium dissipation
improves learning quality and determine the Landauer energy cost of contrastive
learning through physical dynamics.