Generalization from Low- to Moderate-Resolution Spectra with Neural Networks for Stellar Parameter Estimation: A Case Study with DESI
2026-02-16 • Machine Learning
Machine Learning
AI summaryⓘ
The authors studied how to use simple neural networks called MLPs to analyze star light data from different telescopes with different resolutions. They first trained the MLPs on low-resolution data from one survey (LAMOST) and then adapted them to work on medium-resolution data from another survey (DESI). They tested using raw spectra and special compressed representations called embeddings from transformer models. The authors found that MLPs trained directly on the low-resolution data worked well, especially with some fine-tuning, and that embeddings helped only for stars richer in metals. They also discovered that the best way to fine-tune the models depends on which stellar property is being measured.
stellar spectral analysiscross-survey generalizationmultilayer perceptron (MLP)low-resolution spectra (LRS)medium-resolution spectra (MRS)fine-tuningtransformer embeddingsiron abundance ([Fe/H])self-supervised learningspectral foundation models
Authors
Xiaosheng Zhao, Yuan-Sen Ting, Rosemary F. G. Wyse, Alexander S. Szalay, Yang Huang, László Dobos, Tamás Budavári, Viska Wei
Abstract
Cross-survey generalization is a critical challenge in stellar spectral analysis, particularly in cases such as transferring from low- to moderate-resolution surveys. We investigate this problem using pre-trained models, focusing on simple neural networks such as multilayer perceptrons (MLPs), with a case study transferring from LAMOST low-resolution spectra (LRS) to DESI medium-resolution spectra (MRS). Specifically, we pre-train MLPs on either LRS or their embeddings and fine-tune them for application to DESI stellar spectra. We compare MLPs trained directly on spectra with those trained on embeddings derived from transformer-based models (self-supervised foundation models pre-trained for multiple downstream tasks). We also evaluate different fine-tuning strategies, including residual-head adapters, LoRA, and full fine-tuning. We find that MLPs pre-trained on LAMOST LRS achieve strong performance, even without fine-tuning, and that modest fine-tuning with DESI spectra further improves the results. For iron abundance, embeddings from a transformer-based model yield advantages in the metal-rich ([Fe/H] > -1.0) regime, but underperform in the metal-poor regime compared to MLPs trained directly on LRS. We also show that the optimal fine-tuning strategy depends on the specific stellar parameter under consideration. These results highlight that simple pre-trained MLPs can provide competitive cross-survey generalization, while the role of spectral foundation models for cross-survey stellar parameter estimation requires further exploration.