Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry

Neural-network quantum states (NQS) has emerged as a powerful application of quantum-inspired deep learning for variational Monte Carlo methods, offering a competitive alternative to existing techniques for identifying ground states of quantum problems. A significant advancement toward improving the...

Full description

Saved in:
Bibliographic Details
Main Authors: Oliver Knitter, Dan Zhao, James Stokes, Martin Ganahl, Stefan Leichenauer, Shravan Veerapaneni
Format: Article
Language:English
Published: IOP Publishing 2025-01-01
Series:Machine Learning: Science and Technology
Subjects:
Online Access:https://doi.org/10.1088/2632-2153/adcb88
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850178492432384000
author Oliver Knitter
Dan Zhao
James Stokes
Martin Ganahl
Stefan Leichenauer
Shravan Veerapaneni
author_facet Oliver Knitter
Dan Zhao
James Stokes
Martin Ganahl
Stefan Leichenauer
Shravan Veerapaneni
author_sort Oliver Knitter
collection DOAJ
description Neural-network quantum states (NQS) has emerged as a powerful application of quantum-inspired deep learning for variational Monte Carlo methods, offering a competitive alternative to existing techniques for identifying ground states of quantum problems. A significant advancement toward improving the practical scalability of NQS has been the incorporation of autoregressive models, most recently transformers, as variational ansätze. Transformers learn sequence information with greater expressiveness than recurrent models, but at the cost of increased time complexity with respect to sequence length. We explore the use of the retentive network (RetNet), a recurrent alternative to transformers, as an ansatz for solving electronic ground state problems in ab initio quantum chemistry. Unlike transformers, RetNets overcome this time complexity bottleneck by processing data in parallel during training, and recurrently during inference. We give a simple computational cost estimate of the RetNet and directly compare it with similar estimates for transformers, establishing a clear threshold ratio of problem-to-model size past which the RetNet’s time complexity outperforms that of the transformer. Though this efficiency comes at the expense of decreased expressiveness relative to the transformer, we overcome this gap through training strategies that leverage the autoregressive structure of the model—namely, variational neural annealing. Our findings support the RetNet as a means of improving the time complexity of NQS without sacrificing accuracy. We provide further evidence that the ablative improvements of neural annealing extend beyond the RetNet architecture, suggesting it would serve as an effective general training strategy for autoregressive NQS.
format Article
id doaj-art-2e425b81ff98450cb21c331fec80d496
institution OA Journals
issn 2632-2153
language English
publishDate 2025-01-01
publisher IOP Publishing
record_format Article
series Machine Learning: Science and Technology
spelling doaj-art-2e425b81ff98450cb21c331fec80d4962025-08-20T02:18:43ZengIOP PublishingMachine Learning: Science and Technology2632-21532025-01-016202502210.1088/2632-2153/adcb88Retentive neural quantum states: efficient ansätze for ab initio quantum chemistryOliver Knitter0https://orcid.org/0000-0001-9163-943XDan Zhao1James Stokes2Martin Ganahl3Stefan Leichenauer4https://orcid.org/0009-0004-9701-1968Shravan Veerapaneni5https://orcid.org/0000-0002-2294-7233University of Michigan , Ann Arbor, MI, United States of America; SandboxAQ , Palo Alto, CA, United States of AmericaSandboxAQ , Palo Alto, CA, United States of America; New York University , New York, NY, United States of AmericaUniversity of Michigan , Ann Arbor, MI, United States of AmericaSandboxAQ , Palo Alto, CA, United States of AmericaSandboxAQ , Palo Alto, CA, United States of AmericaUniversity of Michigan , Ann Arbor, MI, United States of AmericaNeural-network quantum states (NQS) has emerged as a powerful application of quantum-inspired deep learning for variational Monte Carlo methods, offering a competitive alternative to existing techniques for identifying ground states of quantum problems. A significant advancement toward improving the practical scalability of NQS has been the incorporation of autoregressive models, most recently transformers, as variational ansätze. Transformers learn sequence information with greater expressiveness than recurrent models, but at the cost of increased time complexity with respect to sequence length. We explore the use of the retentive network (RetNet), a recurrent alternative to transformers, as an ansatz for solving electronic ground state problems in ab initio quantum chemistry. Unlike transformers, RetNets overcome this time complexity bottleneck by processing data in parallel during training, and recurrently during inference. We give a simple computational cost estimate of the RetNet and directly compare it with similar estimates for transformers, establishing a clear threshold ratio of problem-to-model size past which the RetNet’s time complexity outperforms that of the transformer. Though this efficiency comes at the expense of decreased expressiveness relative to the transformer, we overcome this gap through training strategies that leverage the autoregressive structure of the model—namely, variational neural annealing. Our findings support the RetNet as a means of improving the time complexity of NQS without sacrificing accuracy. We provide further evidence that the ablative improvements of neural annealing extend beyond the RetNet architecture, suggesting it would serve as an effective general training strategy for autoregressive NQS.https://doi.org/10.1088/2632-2153/adcb88machine learningquantum chemistryneural quantum statesAI for scienceRetNetsvariational neural annealing
spellingShingle Oliver Knitter
Dan Zhao
James Stokes
Martin Ganahl
Stefan Leichenauer
Shravan Veerapaneni
Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry
Machine Learning: Science and Technology
machine learning
quantum chemistry
neural quantum states
AI for science
RetNets
variational neural annealing
title Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry
title_full Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry
title_fullStr Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry
title_full_unstemmed Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry
title_short Retentive neural quantum states: efficient ansätze for ab initio quantum chemistry
title_sort retentive neural quantum states efficient ansatze for ab initio quantum chemistry
topic machine learning
quantum chemistry
neural quantum states
AI for science
RetNets
variational neural annealing
url https://doi.org/10.1088/2632-2153/adcb88
work_keys_str_mv AT oliverknitter retentiveneuralquantumstatesefficientansatzeforabinitioquantumchemistry
AT danzhao retentiveneuralquantumstatesefficientansatzeforabinitioquantumchemistry
AT jamesstokes retentiveneuralquantumstatesefficientansatzeforabinitioquantumchemistry
AT martinganahl retentiveneuralquantumstatesefficientansatzeforabinitioquantumchemistry
AT stefanleichenauer retentiveneuralquantumstatesefficientansatzeforabinitioquantumchemistry
AT shravanveerapaneni retentiveneuralquantumstatesefficientansatzeforabinitioquantumchemistry