Numeric Analytic Continuation via Rational Function Regression (Padé Regression)
POSTER
Abstract
We have developed a simple and natural method to perform numeric analytic conitnuation of quantum Monte Carlo data.
It is based on the Padé approximation, and uses linear regression and bootstrapping statistics to get rid of extreme unstable results, and estimate the error. Unlike maximum entropy method, no prior information is needed here.
Test cases have shown that various spectrum is recovered for relative error as large as 0.1%
Physical reasonings are also explained for its success.
It is based on the Padé approximation, and uses linear regression and bootstrapping statistics to get rid of extreme unstable results, and estimate the error. Unlike maximum entropy method, no prior information is needed here.
Test cases have shown that various spectrum is recovered for relative error as large as 0.1%
Physical reasonings are also explained for its success.
Presenters
-
Jian Wang
physics and astronomy, UCLA, physics and astronomy, University of California, Los Angeles
Authors
-
Jian Wang
physics and astronomy, UCLA, physics and astronomy, University of California, Los Angeles
-
Sudip Chakravarty
physics and astronomy, UCLA, physics and astronomy, University of California, Los Angeles