Matches in SemOpenAlex for { <https://semopenalex.org/work/W3166322571> ?p ?o ?g. }
Showing items 1 to 74 of
74
with 100 items per page.
- W3166322571 endingPage "704" @default.
- W3166322571 startingPage "694" @default.
- W3166322571 abstract "Neural networks (NNs) have been extremely successful across many tasks in machine learning. Quantization of NN weights has become an important topic due to its impact on their energy efficiency, inference time and deployment on hardware. Although post-training quantization is well-studied, training optimal quantized NNs involves combinatorial non-convex optimization problems which appear intractable. In this work, we introduce a convex optimization strategy to train quantized NNs with polynomial activations. Our method leverages hidden convexity in two-layer neural networks from the recent literature, semidefinite lifting, and Grothendieck's identity. Surprisingly, we show that certain quantized NN problems can be solved to global optimality in polynomial-time in all relevant parameters via semidefinite relaxations. We present numerical examples to illustrate the effectiveness of our method." @default.
- W3166322571 created "2021-06-22" @default.
- W3166322571 creator A5001436196 @default.
- W3166322571 creator A5062073113 @default.
- W3166322571 date "2021-07-18" @default.
- W3166322571 modified "2023-09-23" @default.
- W3166322571 title "Training Quantized Neural Networks to Global Optimality via Semidefinite Programming" @default.
- W3166322571 hasPublicationYear "2021" @default.
- W3166322571 type Work @default.
- W3166322571 sameAs 3166322571 @default.
- W3166322571 citedByCount "1" @default.
- W3166322571 countsByYear W31663225712021 @default.
- W3166322571 crossrefType "proceedings-article" @default.
- W3166322571 hasAuthorship W3166322571A5001436196 @default.
- W3166322571 hasAuthorship W3166322571A5062073113 @default.
- W3166322571 hasConcept C101901036 @default.
- W3166322571 hasConcept C106159729 @default.
- W3166322571 hasConcept C112680207 @default.
- W3166322571 hasConcept C11413529 @default.
- W3166322571 hasConcept C126255220 @default.
- W3166322571 hasConcept C137836250 @default.
- W3166322571 hasConcept C154945302 @default.
- W3166322571 hasConcept C157972887 @default.
- W3166322571 hasConcept C162324750 @default.
- W3166322571 hasConcept C2524010 @default.
- W3166322571 hasConcept C2776214188 @default.
- W3166322571 hasConcept C28855332 @default.
- W3166322571 hasConcept C33923547 @default.
- W3166322571 hasConcept C41008148 @default.
- W3166322571 hasConcept C50644808 @default.
- W3166322571 hasConcept C72134830 @default.
- W3166322571 hasConceptScore W3166322571C101901036 @default.
- W3166322571 hasConceptScore W3166322571C106159729 @default.
- W3166322571 hasConceptScore W3166322571C112680207 @default.
- W3166322571 hasConceptScore W3166322571C11413529 @default.
- W3166322571 hasConceptScore W3166322571C126255220 @default.
- W3166322571 hasConceptScore W3166322571C137836250 @default.
- W3166322571 hasConceptScore W3166322571C154945302 @default.
- W3166322571 hasConceptScore W3166322571C157972887 @default.
- W3166322571 hasConceptScore W3166322571C162324750 @default.
- W3166322571 hasConceptScore W3166322571C2524010 @default.
- W3166322571 hasConceptScore W3166322571C2776214188 @default.
- W3166322571 hasConceptScore W3166322571C28855332 @default.
- W3166322571 hasConceptScore W3166322571C33923547 @default.
- W3166322571 hasConceptScore W3166322571C41008148 @default.
- W3166322571 hasConceptScore W3166322571C50644808 @default.
- W3166322571 hasConceptScore W3166322571C72134830 @default.
- W3166322571 hasOpenAccess W3166322571 @default.
- W3166322571 hasRelatedWork W1549275021 @default.
- W3166322571 hasRelatedWork W163939031 @default.
- W3166322571 hasRelatedWork W1982811442 @default.
- W3166322571 hasRelatedWork W1985349699 @default.
- W3166322571 hasRelatedWork W2056157083 @default.
- W3166322571 hasRelatedWork W2092575309 @default.
- W3166322571 hasRelatedWork W2325799880 @default.
- W3166322571 hasRelatedWork W2388965754 @default.
- W3166322571 hasRelatedWork W2542766356 @default.
- W3166322571 hasRelatedWork W2755377039 @default.
- W3166322571 hasRelatedWork W2791251367 @default.
- W3166322571 hasRelatedWork W2794272078 @default.
- W3166322571 hasRelatedWork W279541522 @default.
- W3166322571 hasRelatedWork W2901827714 @default.
- W3166322571 hasRelatedWork W2962928577 @default.
- W3166322571 hasRelatedWork W2992146231 @default.
- W3166322571 hasRelatedWork W3038368057 @default.
- W3166322571 hasRelatedWork W3114439266 @default.
- W3166322571 hasRelatedWork W3117881175 @default.
- W3166322571 hasRelatedWork W3159107992 @default.
- W3166322571 isParatext "false" @default.
- W3166322571 isRetracted "false" @default.
- W3166322571 magId "3166322571" @default.
- W3166322571 workType "article" @default.