Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graph-Based Light-Curve Features for Robust Transient Classification

Jesús D. Petro-Ramos, David J. Ruiz-Morales, D. Sierra-Porta

TL;DR

Results indicate that visibility graphs offer a simple, survey-agnostic bridge between irregular photometric time series and standard classifiers, yielding competitive multiclass performance without bespoke deep architectures.

Abstract

We investigate graph-based representations of astronomical light curves for transient classification on the MANTRA benchmark. Each series is mapped to three visibility-graph views-horizontal (HVG), directed (DHVG), and weighted (W-HVG)-from which we extract compact, length-aware network descriptors (degree/strength moments, clustering and motifs, assortativity, path/efficiency, and spectral summaries). Using stratified five-fold validation and tree-based learners, the best configuration (LightGBM with HVG+DHVG+W-HVG features) attains a macro-F1 of 0.622 +/- 0.010 and accuracy of 0.661 +/- 0.010, improving over the MANTRA baseline (F1 macro = 0.528). Ablations show that weighted contrasts and directed asymmetry contribute complementary gains to undirected topology. Per-class analysis highlights strong performance for CV, HPM, and Non-Tr., with residual confusions concentrated in the AGN-Blazar-SN block. These results indicate that visibility graphs offer a simple, survey-agnostic bridge between irregular photometric time series and standard classifiers, yielding competitive multiclass performance without bespoke deep architectures. We release code and feature definitions to facilitate reproducibility and future extensions.

Graph-Based Light-Curve Features for Robust Transient Classification

TL;DR

Results indicate that visibility graphs offer a simple, survey-agnostic bridge between irregular photometric time series and standard classifiers, yielding competitive multiclass performance without bespoke deep architectures.

Abstract

We investigate graph-based representations of astronomical light curves for transient classification on the MANTRA benchmark. Each series is mapped to three visibility-graph views-horizontal (HVG), directed (DHVG), and weighted (W-HVG)-from which we extract compact, length-aware network descriptors (degree/strength moments, clustering and motifs, assortativity, path/efficiency, and spectral summaries). Using stratified five-fold validation and tree-based learners, the best configuration (LightGBM with HVG+DHVG+W-HVG features) attains a macro-F1 of 0.622 +/- 0.010 and accuracy of 0.661 +/- 0.010, improving over the MANTRA baseline (F1 macro = 0.528). Ablations show that weighted contrasts and directed asymmetry contribute complementary gains to undirected topology. Per-class analysis highlights strong performance for CV, HPM, and Non-Tr., with residual confusions concentrated in the AGN-Blazar-SN block. These results indicate that visibility graphs offer a simple, survey-agnostic bridge between irregular photometric time series and standard classifiers, yielding competitive multiclass performance without bespoke deep architectures. We release code and feature definitions to facilitate reproducibility and future extensions.
Paper Structure (10 sections, 4 equations, 3 figures, 2 tables)

This paper contains 10 sections, 4 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Data Description
  3. Methods
  4. Feature generation: visibility graphs and network descriptors
  5. Network features
  6. Learning algorithms
  7. Evaluation protocol
  8. Results
  9. Conclusions
  10. Data and Code Availability

Figures (3)

  • Figure 1: Out-of-fold predictions for the best configuration (LGBM + HVG+DHVG+W-HVG) on the eight-class MANTRA task.
  • Figure 2: One-vs-rest PR–AUC per class (OOF). Higher values indicate better separability against the rest of the catalog.
  • Figure 3: Model-based feature importances aggregated across folds for the best system. Bars show mean importance; whiskers show the standard deviation across folds.