Back to projects
Research Projects
01 High-redshift black holes and AGNs 02 High-redshift galaxies and star formation 03 Active Artificial Intelligence for astronomy

Project 03 / Survey-scale inference

Artificial Intelligence for Astronomy

This project uses JWST-calibrated machine learning to classify galaxies, estimate physical parameters, and develop methods for large-area surveys.

SHAPE methodology image

Works

Machine-learning methods for astronomy.

SHAPE SOM-SED hybrid approach visual

Published / JWST-calibrated estimator

Efficient galaxy parameter estimation leveraging JWST observations

Author list: Zihao Wang, Tao Wang, Ke Xu, Hanwen Sun, Ruining Tian, Qi Hao

SHAPE uses the largest and deepest JWST/NIRCam and MIRI surveys from PRIMER to calibrate efficient galaxy parameter estimates, focusing on galaxies at z ~ 1.5-2.5 as a methodological test.

  • Replaces default SOM weights with stacked SEDs from each cell, extending the model to other photometric catalogs.
  • Uses an SED library to apply the JWST-calibrated model to COSMOS2020.
  • Produces stellar mass and star-formation-rate estimates that agree with extended-photometry SED fitting.
  • Improves SFR estimation under identical photometric constraints while remaining computationally efficient.
  • Provides a route toward survey-scale inference for CSST, Euclid, and large-area deep-field catalogs.