Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035170382> ?p ?o ?g. }
- W3035170382 abstract "The state-of-the art machine learning approach to training deep neural networks, backpropagation, is implausible for real neural networks: neurons need to know their outgoing weights; training alternates between a bottom-up forward pass (computation) and a top-down backward pass (learning); and the algorithm often needs precise labels of many data points. Biologically plausible approximations to backpropagation, such as feedback alignment, solve the weight transport problem, but not the other two. Thus, fully biologically plausible learning rules have so far remained elusive. Here we present a family of learning rules that does not suffer from any of these problems. It is motivated by the information bottleneck principle (extended with kernel methods), in which networks learn to compress the input as much as possible without sacrificing prediction of the output. The resulting rules have a 3-factor Hebbian structure: they require pre- and post-synaptic firing rates and an error signal - the third factor - consisting of a global teaching signal and a layer-specific term, both available without a top-down pass. They do not require precise labels; instead, they rely on the similarity between pairs of desired outputs. Moreover, to obtain good performance on hard problems and retain biological plausibility, our rules need divisive normalization - a known feature of biological networks. Finally, simulations show that our rules perform nearly as well as backpropagation on image classification tasks." @default.
- W3035170382 created "2020-06-19" @default.
- W3035170382 creator A5025338734 @default.
- W3035170382 creator A5027156456 @default.
- W3035170382 date "2020-06-12" @default.
- W3035170382 modified "2023-09-27" @default.
- W3035170382 title "Kernelized information bottleneck leads to biologically plausible 3-factor Hebbian learning in deep networks" @default.
- W3035170382 cites W1492899839 @default.
- W3035170382 cites W1522301498 @default.
- W3035170382 cites W1606458877 @default.
- W3035170382 cites W1638081485 @default.
- W3035170382 cites W1686946872 @default.
- W3035170382 cites W1988013646 @default.
- W3035170382 cites W2062262801 @default.
- W3035170382 cites W2134664762 @default.
- W3035170382 cites W2138260451 @default.
- W3035170382 cites W2527798464 @default.
- W3035170382 cites W2552737632 @default.
- W3035170382 cites W2557579533 @default.
- W3035170382 cites W2559752379 @default.
- W3035170382 cites W2593634001 @default.
- W3035170382 cites W2735683573 @default.
- W3035170382 cites W2750384547 @default.
- W3035170382 cites W2771657877 @default.
- W3035170382 cites W2786738752 @default.
- W3035170382 cites W2810701348 @default.
- W3035170382 cites W2842511635 @default.
- W3035170382 cites W2899171226 @default.
- W3035170382 cites W2902986194 @default.
- W3035170382 cites W2903723132 @default.
- W3035170382 cites W2911586496 @default.
- W3035170382 cites W2944546415 @default.
- W3035170382 cites W2949117887 @default.
- W3035170382 cites W2949405272 @default.
- W3035170382 cites W2950541952 @default.
- W3035170382 cites W2951720195 @default.
- W3035170382 cites W2961192671 @default.
- W3035170382 cites W2963047948 @default.
- W3035170382 cites W2963440770 @default.
- W3035170382 cites W2963454111 @default.
- W3035170382 cites W2964115671 @default.
- W3035170382 cites W2964319207 @default.
- W3035170382 cites W2970674713 @default.
- W3035170382 cites W2978368159 @default.
- W3035170382 cites W2990583358 @default.
- W3035170382 cites W2998249317 @default.
- W3035170382 cites W3005680577 @default.
- W3035170382 cites W3008951969 @default.
- W3035170382 cites W3049154595 @default.
- W3035170382 cites W3105516366 @default.
- W3035170382 cites W3118608800 @default.
- W3035170382 cites W3130377169 @default.
- W3035170382 hasPublicationYear "2020" @default.
- W3035170382 type Work @default.
- W3035170382 sameAs 3035170382 @default.
- W3035170382 citedByCount "5" @default.
- W3035170382 countsByYear W30351703822020 @default.
- W3035170382 countsByYear W30351703822021 @default.
- W3035170382 crossrefType "posted-content" @default.
- W3035170382 hasAuthorship W3035170382A5025338734 @default.
- W3035170382 hasAuthorship W3035170382A5027156456 @default.
- W3035170382 hasConcept C108583219 @default.
- W3035170382 hasConcept C111437709 @default.
- W3035170382 hasConcept C114614502 @default.
- W3035170382 hasConcept C117765406 @default.
- W3035170382 hasConcept C119857082 @default.
- W3035170382 hasConcept C120822770 @default.
- W3035170382 hasConcept C136886441 @default.
- W3035170382 hasConcept C138885662 @default.
- W3035170382 hasConcept C144024400 @default.
- W3035170382 hasConcept C149635348 @default.
- W3035170382 hasConcept C153180895 @default.
- W3035170382 hasConcept C154945302 @default.
- W3035170382 hasConcept C155032097 @default.
- W3035170382 hasConcept C17061570 @default.
- W3035170382 hasConcept C19165224 @default.
- W3035170382 hasConcept C2776401178 @default.
- W3035170382 hasConcept C2779127903 @default.
- W3035170382 hasConcept C2780513914 @default.
- W3035170382 hasConcept C33923547 @default.
- W3035170382 hasConcept C41008148 @default.
- W3035170382 hasConcept C41895202 @default.
- W3035170382 hasConcept C50644808 @default.
- W3035170382 hasConcept C60008888 @default.
- W3035170382 hasConcept C73555534 @default.
- W3035170382 hasConcept C74193536 @default.
- W3035170382 hasConcept C97108695 @default.
- W3035170382 hasConceptScore W3035170382C108583219 @default.
- W3035170382 hasConceptScore W3035170382C111437709 @default.
- W3035170382 hasConceptScore W3035170382C114614502 @default.
- W3035170382 hasConceptScore W3035170382C117765406 @default.
- W3035170382 hasConceptScore W3035170382C119857082 @default.
- W3035170382 hasConceptScore W3035170382C120822770 @default.
- W3035170382 hasConceptScore W3035170382C136886441 @default.
- W3035170382 hasConceptScore W3035170382C138885662 @default.
- W3035170382 hasConceptScore W3035170382C144024400 @default.
- W3035170382 hasConceptScore W3035170382C149635348 @default.
- W3035170382 hasConceptScore W3035170382C153180895 @default.
- W3035170382 hasConceptScore W3035170382C154945302 @default.
- W3035170382 hasConceptScore W3035170382C155032097 @default.