specsrbench

Benchmarking deep learning against classical deconvolution for galaxy spectral super-resolution.

Seven classical deconvolution methods — cubic interpolation, Wiener, Tikhonov, Wiener + total variation, Richardson–Lucy, wavelet-sparse FISTA, and a redshift-informed matched filter — scored against the SR2 deep-learning pipeline of specsr on 572 held-out JWST/NIRSpec prism spectra from JADES.

The result, stated plainly

At the pixel level, no method beats cubic interpolation.

SR2 leads the raw mean-absolute-error table by 30%, and does it by producing a spectrum at 0.54 of the reference’s amplitude. On a scale-free metric all nine methods land within 1.2% of each other and SR2 ranks eighth of nine.

What survives is more interesting than a leaderboard: SR2 exceeds the reference’s own line signal-to-noise on all four diagnostic lines while recovering only 36–53% of true line amplitudes, at false-detection rates of 0.30 (Hβ) and 0.44 ([O II]) against ≤ 0.09 for anything classical.

Quickstart

specsrbench paths           # where it will look for inputs
specsrbench figures all     # the six paper figures
specsrbench build all       # the cache they read, from JADES DR4 + the Hub

Or from Python:

from specsrbench.data import load_cache
from specsrbench.metrics import mae, mae_scalefree, std_ratio

cache = load_cache()
truth, mask = cache.x_high, cache.valid

for key, recon in cache.arrays.items():
    print(f"{key:9s} MAE {mae(recon, truth, mask):.4f}"
          f"  scale-free {mae_scalefree(recon, truth, mask):.4f}"
          f"  amplitude {std_ratio(recon, truth, mask):.3f}")

Reporting mae without std_ratio beside it is how this project twice stated a conclusion the scale-free number reverses. Four ways mean absolute error lies is the long version of why.

Reference

Paper 2 is in preparation. For the deep-learning models it benchmarks, see

Haghjoo, A., Hemmati, S., Mobasher, B., et al. Learning to See Sharper: A Physics-Informed Artificial Intelligence Framework for Super-Resolving Galaxy Spectra. arXiv:2603.18357

Guides

Reference