aiqbenchInformații despre AI & calcul cuantic

...

Română · beta. Traducerea a fost făcută automat și este încă în curs de revizuire. Citește pagina în engleză

perspectiva aiqbenchCercetare AI

This research demonstrates that incorporating parity labels into neural network architectures is essential for enforcing fundamental physical symmetries in materials discovery models. By explicitly encoding how features transform under spatial inversion, this approach reduces physically impossible predictions by six orders of magnitude compared to data-driven training alone. The findings highlight a critical architectural requirement for ensuring machine learning models adhere to the laws of physics when predicting crystal properties.

Surse care acoperăQuantum Zeitgeist, Quantum Computing Report

Ce este aceastaLectura unui model asupra articolului și a acoperirii conexe, scrisă pentru aiqbench. Nu sunt cuvintele publicației.

Design Choice Limits Impossible Predictions to Less Than One Percent of Crystals

Quantum Zeitgeist16 septembrie 20265 min de cititArticolul și perspectiva sunt în engleză.

A single architectural detail determines whether machine learning models correctly predict fundamental physical laws governing materials behaviour. Parity-labelled architectures achieved results indistinguishable from zero, whereas comparable models predicted impossible responses for 90 to 96 per cent of tested crystal structures, a difference of six orders of magnitude. This reveals that enforcing mathematical constraints via parity labels is more effective than attempting to learn symmetry through training data alone.

Machine learning models used for materials discovery frequently predicted physically impossible properties until now. Can Polat from Texas A&M University and colleagues have shown that accurate prediction of symmetry-forbidden responses relies on a key architectural detail: parity labels within their features. They identified a fundamental flaw impacting machine learning’s ability to accurately design new materials; without specific construction with ‘parity labels’, models often disregard essential physical symmetries. These parity labels, information about how features reflect across space, sharply improve prediction accuracy by reducing incorrect…

Extras afișat conform termenilor RSS ai publicației · 84 cuvinte · textul integral și figurile sunt pe site-ul publicației

Citește pe Quantum Zeitgeist ↗Întreabă despre acest articol