Gradient-estimator design overcomes trainability barriers in neural-network-based variational optimization (Q7353464)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
No description defined
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Gradient-estimator design overcomes trainability barriers in neural-network-based variational optimization |
No description defined |
Statements
Gradient-estimator design overcomes trainability barriers in neural-network-based variational optimization (English)
0 references
16 September 2026
0 references
cond-mat.dis-nn
0 references
cond-mat.str-el
0 references
cs.LG
0 references
physics.chem-ph
0 references
physics.comp-ph
0 references
Yi-Ran Xue
0 references
Rui Wang
0 references
Baigeng Wang
0 references
Chenan Wei
0 references
We derive a direct unbiased gradient estimator and an adaptive minimum-variance phase (AMVP) estimator to overcome trainability barriers in neural-network-based variational optimization, achieving compact networks with over an order of magnitude less GPU time while matching or surpassing density matrix renormalization group accuracy and chemical accuracy in challenging molecular systems.
0 references
Neural Network
0 references
Gradient Estimation
0 references
Variational Optimization
0 references
1 reference
1 reference
