Mitigating variance shrinkage in multi-task galaxy property inference from imaging data via statistical manifold alignment

Inferring photometric redshifts ($z$) and key stellar population properties ($\log(M_*/\mathrm{M}_\odot)$, $D_n4000$, and $\mathrm{H}δ_A$) directly from multi-band imaging is increasingly critical for spectroscopically limited large surveys such as LSST and CSST. However, conventional point-wise loss functions (e.g., mean squared error or Huber loss) employed in deep neural networks often induce ``variance shrinkage''---a regression-to-the-mean effect that contracts predicted variances, distorts non-linear physical manifolds (such as bimodal distributions), and truncates heavy-tail populations. Here, we present ViTMultiTask, a vision transformer framework that incorporates a multi-scale maximum mean discrepancy (MMD) loss to enforce higher-order joint distribution alignment in reproducing kernel Hilbert space (RKHS). Trained on 83,846 galaxies from Subaru HSC five-band ($g,r,i,z,y$) imaging cross-matched with DESI Data Release 1 Bright Galaxy Survey spectra, the model achieves high point-estimate precision on an independent test set of 5,000 galaxies: $σ_{\mathrm{NMAD}} = 0.020$ for $z$, $0.169\,\mathrm{dex}$ for $\log(M_*/\mathrm{M}_\odot)$, $0.073$ for $D_n4000$, and $1.185\,\text{Å}$ for $\mathrm{H}δ_A$. Ablation experiments demonstrate that incorporating the MMD loss effectively mitigates variance shrinkage and more accurately reproduces the observed spectroscopic parameter distributions (including bimodal structures and heavy tails), while maintaining point-estimate accuracy statistically consistent with the baseline (RMSE changes $\le 3.5\%$) and catastrophic outlier rates fully comparable to the baseline. This kernel-based manifold alignment offers a statistically grounded and practical paradigm for comprehensive galaxy property inference in the deep imaging era.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-12
DOI
https://doi.org/10.48550/arxiv.2611.16598
Primary Topic
Galaxies: Formation, Evolution, Phenomena
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Mitigating variance shrinkage in multi-task galaxy property inference from imaging data via statistical manifold alignment

Jianzhen Chen, Zhijian Luo
arXiv (Cornell University)
Galaxies: Formation, Evolution, Phenomena
preprint

Mitigating variance shrinkage in multi-task galaxy property inference from imaging data via statistical manifold alignment

Jianzhen Chen, Zhijian Luo
preprint en

Abstract

Inferring photometric redshifts ($z$) and key stellar population properties ($\log(M_*/\mathrm{M}_\odot)$, $D_n4000$, and $\mathrm{H}δ_A$) directly from multi-band imaging is increasingly critical for spectroscopically limited large surveys such as LSST and CSST. However, conventional point-wise loss functions (e.g., mean squared error or Huber loss) employed in deep neural networks often induce ``variance shrinkage''---a regression-to-the-mean effect that contracts predicted variances, distorts non-linear physical manifolds (such as bimodal distributions), and truncates heavy-tail populations. Here, we present ViTMultiTask, a vision transformer framework that incorporates a multi-scale maximum mean discrepancy (MMD) loss to enforce higher-order joint distribution alignment in reproducing kernel Hilbert space (RKHS). Trained on 83,846 galaxies from Subaru HSC five-band ($g,r,i,z,y$) imaging cross-matched with DESI Data Release 1 Bright Galaxy Survey spectra, the model achieves high point-estimate precision on an independent test set of 5,000 galaxies: $σ_{\mathrm{NMAD}} = 0.020$ for $z$, $0.169\,\mathrm{dex}$ for $\log(M_*/\mathrm{M}_\odot)$, $0.073$ for $D_n4000$, and $1.185\,\text{Å}$ for $\mathrm{H}δ_A$. Ablation experiments demonstrate that incorporating the MMD loss effectively mitigates variance shrinkage and more accurately reproduces the observed spectroscopic parameter distributions (including bimodal structures and heavy tails), while maintaining point-estimate accuracy statistically consistent with the baseline (RMSE changes $\le 3.5\%$) and catastrophic outlier rates fully comparable to the baseline. This kernel-based manifold alignment offers a statistically grounded and practical paradigm for comprehensive galaxy property inference in the deep imaging era.

arXiv (Cornell University)
Galaxies: Formation, Evolution, Phenomena
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon