SCoPe: ZTF Source Classification Project

scope-ml uses machine learning to classify variable star light curves from the Zwicky Transient Facility (ZTF) and the Vera C. Rubin Observatory (LSST).

What SCoPe does

• Feature generation from light curves: period-finding (Conditional Entropy, Analysis of Variance, Lomb-Scargle, FPW), Fourier decomposition, and statistical features via the periodfind library
• Classification using XGBoost and Deep Neural Network binary classifiers trained on ~80,000 manually-labeled sources
• Integration with the Fritz transient broker for uploading/downloading classifications
• Scalable processing via SLURM for large-scale feature generation and inference across ZTF fields

Supported data sources

• ZTF light curves via the Kowalski database
• Rubin DP1 forced photometry via TAP API or local parquet files
• External catalogs: Gaia EDR3, AllWISE, Pan-STARRS1

Quick links

• Installation -- get started with pip install scope-ml
• Quick Start -- train your first classifier in minutes
• Feature Generation -- generate features from ZTF light curves
• Rubin DP1 -- process Rubin Data Preview 1 data
• Field Guide -- learn about the source classes SCoPe identifies
• CLI Reference -- all available commands

Funding

We gratefully acknowledge previous and current support from the U.S. National Science Foundation (NSF) Harnessing the Data Revolution (HDR) Institute for Accelerated AI Algorithms for Data-Driven Discovery (A3D3) under Cooperative Agreement No. PHY-2117997.