SciPost Submission Page
SmoQyDEAC.jl: A differential evolution package for the analytic continuation of imaginary time correlation functions
by James Neuhaus, Nathan S. Nichols, Debshikha Banerjee, Benjamin Cohen-Stead, Thomas A. Maier, Adrian Del Maestro, Steven Johnston
This is not the latest submitted version.
This Submission thread is now published as
Submission summary
| Authors (as registered SciPost users): | James Neuhaus |
| Submission information | |
|---|---|
| Preprint Link: | https://arxiv.org/abs/2407.04568v1 (pdf) |
| Code repository: | https://github.com/SmoQySuite/SmoQyDEAC.jl |
| Code version: | v1.0.0 |
| Code license: | Creative Commons Attribution 4.0 International (CC BY 4.0) License |
| Data repository: | https://zenodo.org/records/10407525 |
| Date submitted: | 2024-07-08 16:17 |
| Submitted by: | Neuhaus, James |
| Submitted to: | SciPost Physics Codebases |
| Ontological classification | |
|---|---|
| Academic field: | Physics |
| Specialties: |
|
| Approach: | Computational |
Abstract
We introduce the SmoQyDEAC.jl package, a Julia implementation of the Differential Evolution Analytic Continuation (DEAC) algorithm [N. S. Nichols et al., Phys. Rev. E 106, 025312 (2022)] for analytically continuing noisy imaginary time correlation functions to the real frequency axis. Our implementation supports fermionic and bosonic correlation functions on either the imaginary time or Matsubara frequency axes, and treatment of the covariance error in the input data. This paper presents an overview of the DEAC algorithm and the features implemented in the SmoQyDEAC.jl. It also provides detailed benchmarks of the package's output against the popular maximum entropy and stochastic analytic continuation methods. The code for this package can be downloaded from our GitHub repository at https://github.com/SmoQySuite/SmoQyDEAC.jl or installed using the Julia package manager. The online documentation, including examples, can be accessed at https://smoqysuite.github.io/SmoQyDEAC.jl/stable/.
Current status:
Reports on this Submission
Report #1 by Anonymous (Referee 1) on 2024-9-14 (Invited Report)
- Cite as: Anonymous, Report on arXiv:2407.04568v1, delivered 2024-09-14, doi: 10.21468/SciPost.Report.9764
Strengths
1. clear documentation for API and examples
2. well-formatted manuscript
Weaknesses
NA
Report
The authors present a Julia package that implements the Differential Evolution Analytic Continuation (DEAC) method, a technique recently proposed for continuing noisy correlation functions on the imaginary axis. This software package could benefit the condensed matter community, especially those lacking computational expertise necessary for performing analytical continuation on simulation data.
The manuscript is well-written and the documentation is quite clear.
The authors provide a handful benchmark against other well-established methods, which should help users become familiar with both the package and the underlying algorithm.
I only have one minor question before recommending this manuscript for publication in SciPost.
Requested changes
1. The DEAC results often exhibit oscillating curves, as shown in Figures 2, 3, and 4, where the authors attribute these oscillations to noise from Monte Carlo (MC) sampling. However, I find it somewhat confusing to determine whether the oscillations stem from insufficient MC sampling or from overfitting. In other words, are all the oscillations in Figures 2, 3, and 4 suppressible as the number of MC sampling increases? Could the authors demonstrate the convergence with respect to the number of runs for one or more of the selected test systems? Such an analysis would be highly instructive, providing users with clear guidance on how to assess the quality of their DEAC results.
Recommendation
Ask for minor revision
Author: James Neuhaus on 2024-10-02 [id 4823]
(in reply to Report 1 on 2024-09-14)We thank the Reviewer for their time and for their support for publication.
We believe that the oscillations in the DEAC data referenced by the Reviewer are the rapid
ones shown in each case due to the statistical noise introduced by sampling the population. One can
systematically decrease this noise level by averaging over more genomes. To demonstrate this, we have
added a new appendix to the paper that shows the evolution of the predicted spectra shown in Fig. 2e
as the total number of genomes increases. We have also added another appendix that discusses some
common spectral features we have encountered when we overfit a given spectrum. Attached are the two new figures which are included in the new submission. A noise vs number of genomes plot is on the left hand side, and a plot of target fitness is on the right hand side. Full explanations are included in the two new appendices.
Thank you.
Attachment: