Matches in SemOpenAlex for { <https://semopenalex.org/work/W2973506077> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W2973506077 abstract "Hardware architectures composed of resistive cross-point device arrays can provide significant power and speed benefits for deep neural network training workloads using stochastic gradient descent (SGD) and backpropagation (BP) algorithm. The training accuracy on this imminent analog hardware however strongly depends on the switching characteristics of the cross-point elements. One of the key requirements is that these resistive devices must change conductance in a symmetrical fashion when subjected to positive or negative pulse stimuli. Here, we present a new training algorithm, so-called the Tiki-Taka algorithm, that eliminates this stringent symmetry requirement. We show that device asymmetry introduces an unintentional implicit cost term into the SGD algorithm, whereas in the Tiki-Taka algorithm a coupled dynamical system simultaneously minimizes the original objective function of the neural network and the unintentional cost term due to device asymmetry in a self-consistent fashion. We tested the validity of this new algorithm on a range of network architectures such as fully connected, convolutional and LSTM networks. Simulation results on these various networks show that whatever accuracy is achieved using the conventional SGD algorithm with symmetric (ideal) device switching characteristics the same accuracy is also achieved using the Tiki-Taka algorithm with non-symmetric (non-ideal) device switching characteristics. Moreover, all the operations performed on the arrays are still parallel and therefore the implementation cost of this new algorithm on array architectures is minimal; and it maintains the aforementioned power and speed benefits. These algorithmic improvements are crucial to relax the material specification and to realize technologically viable resistive crossbar arrays that outperform digital accelerators for similar training tasks." @default.
- W2973506077 created "2019-09-26" @default.
- W2973506077 creator A5025512079 @default.
- W2973506077 creator A5073094974 @default.
- W2973506077 date "2019-09-17" @default.
- W2973506077 modified "2023-09-27" @default.
- W2973506077 title "Algorithm for Training Neural Networks on Resistive Device Arrays" @default.
- W2973506077 cites W1542981317 @default.
- W2973506077 cites W1841592590 @default.
- W2973506077 cites W1969627901 @default.
- W2973506077 cites W2158899491 @default.
- W2973506077 cites W2160815625 @default.
- W2973506077 cites W2162390675 @default.
- W2973506077 cites W2163605009 @default.
- W2973506077 cites W2248832573 @default.
- W2973506077 cites W2259472270 @default.
- W2973506077 cites W2307193480 @default.
- W2973506077 cites W2525778437 @default.
- W2973506077 cites W2552299751 @default.
- W2973506077 cites W2560615381 @default.
- W2973506077 cites W2603646652 @default.
- W2973506077 cites W2725513608 @default.
- W2973506077 cites W2896122000 @default.
- W2973506077 cites W2912495531 @default.
- W2973506077 cites W2919115771 @default.
- W2973506077 cites W2949650786 @default.
- W2973506077 cites W2963059095 @default.
- W2973506077 cites W2963387357 @default.
- W2973506077 hasPublicationYear "2019" @default.
- W2973506077 type Work @default.
- W2973506077 sameAs 2973506077 @default.
- W2973506077 citedByCount "1" @default.
- W2973506077 countsByYear W29735060772019 @default.
- W2973506077 crossrefType "posted-content" @default.
- W2973506077 hasAuthorship W2973506077A5025512079 @default.
- W2973506077 hasAuthorship W2973506077A5073094974 @default.
- W2973506077 hasConcept C11413529 @default.
- W2973506077 hasConcept C121332964 @default.
- W2973506077 hasConcept C127413603 @default.
- W2973506077 hasConcept C14036430 @default.
- W2973506077 hasConcept C146978453 @default.
- W2973506077 hasConcept C154945302 @default.
- W2973506077 hasConcept C163258240 @default.
- W2973506077 hasConcept C204323151 @default.
- W2973506077 hasConcept C31972630 @default.
- W2973506077 hasConcept C41008148 @default.
- W2973506077 hasConcept C50644808 @default.
- W2973506077 hasConcept C62520636 @default.
- W2973506077 hasConcept C6899612 @default.
- W2973506077 hasConcept C78458016 @default.
- W2973506077 hasConcept C81363708 @default.
- W2973506077 hasConcept C86803240 @default.
- W2973506077 hasConceptScore W2973506077C11413529 @default.
- W2973506077 hasConceptScore W2973506077C121332964 @default.
- W2973506077 hasConceptScore W2973506077C127413603 @default.
- W2973506077 hasConceptScore W2973506077C14036430 @default.
- W2973506077 hasConceptScore W2973506077C146978453 @default.
- W2973506077 hasConceptScore W2973506077C154945302 @default.
- W2973506077 hasConceptScore W2973506077C163258240 @default.
- W2973506077 hasConceptScore W2973506077C204323151 @default.
- W2973506077 hasConceptScore W2973506077C31972630 @default.
- W2973506077 hasConceptScore W2973506077C41008148 @default.
- W2973506077 hasConceptScore W2973506077C50644808 @default.
- W2973506077 hasConceptScore W2973506077C62520636 @default.
- W2973506077 hasConceptScore W2973506077C6899612 @default.
- W2973506077 hasConceptScore W2973506077C78458016 @default.
- W2973506077 hasConceptScore W2973506077C81363708 @default.
- W2973506077 hasConceptScore W2973506077C86803240 @default.
- W2973506077 hasLocation W29735060771 @default.
- W2973506077 hasOpenAccess W2973506077 @default.
- W2973506077 hasPrimaryLocation W29735060771 @default.
- W2973506077 hasRelatedWork W108740333 @default.
- W2973506077 hasRelatedWork W2090748607 @default.
- W2973506077 hasRelatedWork W2402098947 @default.
- W2973506077 hasRelatedWork W2414581469 @default.
- W2973506077 hasRelatedWork W2512985564 @default.
- W2973506077 hasRelatedWork W2737173947 @default.
- W2973506077 hasRelatedWork W2810376858 @default.
- W2973506077 hasRelatedWork W2891230170 @default.
- W2973506077 hasRelatedWork W2927803337 @default.
- W2973506077 hasRelatedWork W2963325954 @default.
- W2973506077 hasRelatedWork W2970604398 @default.
- W2973506077 hasRelatedWork W2973313988 @default.
- W2973506077 hasRelatedWork W3040209743 @default.
- W2973506077 hasRelatedWork W3046842446 @default.
- W2973506077 hasRelatedWork W3117263620 @default.
- W2973506077 hasRelatedWork W3134847006 @default.
- W2973506077 hasRelatedWork W3175345297 @default.
- W2973506077 hasRelatedWork W3186222796 @default.
- W2973506077 hasRelatedWork W3198405160 @default.
- W2973506077 hasRelatedWork W3116241656 @default.
- W2973506077 isParatext "false" @default.
- W2973506077 isRetracted "false" @default.
- W2973506077 magId "2973506077" @default.
- W2973506077 workType "article" @default.