ORACLE: A Real-Time, Hierarchical, Deep-Learning Photometric Classifier for the LSST

2025-01-02

AI Summary

To address the challenge of early-stage (particularly single-epoch) classification of transients and variables in the Legacy Survey of Space and Time (LSST), this paper introduces the first real-time, context-aware hierarchical deep learning classifier tailored for LSST. Methodologically, we design a GRU-based light-curve encoder and integrate multi-dimensional contextual features—including host-galaxy redshift and positional offsets—while proposing an observation-driven hierarchical cross-entropy loss to enable context-sensitive fine-grained classification. Key contributions include: (1) the first hierarchical GRU architecture coupled with a hierarchical loss function; (2) macro-averaged accuracies of 0.96 for transient/variable binary classification and 0.83 for 19-class fine-grained classification using only the first-night single-epoch photometry; and (3) binary classification accuracy exceeding 0.99 within 64 days—significantly earlier than existing methods—enabling ultra-early scientific response for LSST.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

We present ORACLE, the first hierarchical deep-learning model for real-time, context-aware classification of transient and variable astrophysical phenomena. ORACLE is a recurrent neural network with Gated Recurrent Units (GRUs), and has been trained using a custom hierarchical cross-entropy loss function to provide high-confidence classifications along an observationally-driven taxonomy with as little as a single photometric observation. Contextual information for each object, including host galaxy photometric redshift, offset, ellipticity and brightness, is concatenated to the light curve embedding and used to make a final prediction. Training on $sim$0.5M events from the Extended LSST Astronomical Time-Series Classification Challenge, we achieve a top-level (Transient vs Variable) macro-averaged precision of 0.96 using only 1 day of photometric observations after the first detection in addition to contextual information, for each event; this increases to $>$0.99 once 64 days of the light curve has been obtained, and 0.83 at 1024 days after first detection for 19-way classification (including supernova sub-types, active galactic nuclei, variable stars, microlensing events, and kilonovae). We also compare ORACLE with other state-of-the-art classifiers and report comparable performance for the 19-way classification task, in addition to delivering accurate top-level classifications much earlier. The code and model weights used in this work are publicly available at our associated GitHub repository (https://github.com/uiucsn/ELAsTiCC-Classification).

Problem

• Real-time Classification
• Environmental Sensitivity
• Limited Observation

Innovation

• ORACLE
• GRUs
• context-aware deep learning