On Representing Electronic Wave Functions with Sign Equivariant Neural Networks
Stats
Recent neural networks demonstrated impressively accurate approximations of electronic ground-state wave functions.
In classical quantum chemistry, one may typically require thousands to millions of determinants to capture electronic correlations correctly.
We aim to reduce the classical large number of determinants via neural networks.
The distribution of wave function amplitudes varies by several orders of magnitude.
Quotes
"Recent works improved these Ans¨atze by replacing the so-called orbital functions ϕi in Equation 2 by neural networks."
"We conclude with neither theoretical nor empirical advantages of sign equivariant functions for representing electronic wave functions within the evaluation of this work."
"While our experimental results show that such odd functions combined with classical determinants can yield better results on small structures, optimizing such odd functions proves difficult."