AI summaryⓘ
The authors developed a way to teach computers to tell apart real space events (like exploding stars) from false signals without needing people to label lots of examples. They used fake, simulated events mixed with real survey data that mostly has false signals to train a special model that can handle noisy and mixed-up labels. Their approach also estimates how confident the model is when making decisions, using a new hybrid method that works well and is less expensive to run. They tested their method and found it works reliably even with tough data, and it can also classify different kinds of real events based on their brightness over time. This method can be reused easily for future space surveys by simply rerunning their training using simulated injections.
Real-Bogus classificationTime-domain surveysSimulated transient injectionsLabel noiseAsymmetric co-teachingUncertainty quantificationMC dropoutDeep ensemblesLight curvesLatent-space visualization
Authors
Raphaël Bonnet-Guerrini, Bruno Sanchez, Dominique Fouchez, Benjamin Racine, Maya Guy, Mariam Sabalbal, Manal Yassine, Vincenzo Piuri
Abstract
Time-domain surveys generate many transient candidates, making Real-Bogus classification a critical step in automated discovery pipelines. Reliable labels are costly, while community labels can be noisy and survey-dependent. We aim to develop a Real-Bogus classification framework that can be trained without human-labeled data using injected transients and bogus-dominated survey data, remains robust under strong class contamination, and provides calibrated uncertainty quantification. We combine simulated transient injections with a contaminated survey class and train a dual-network model using asymmetric co-teaching for classes with different label-noise levels. We evaluate performance on a benchmark subset and analyze the learned representation with latent-space visualization tools. For uncertainty quantification (UQ), we compare MC dropout and deep ensembles and propose a low-cost hybrid strategy that exploits the dual-network setting to improve calibration. We extend the evaluation to the light-curve domain to assess recovery of light-curve classes. The method achieves strong Real-Bogus performance on the labeled subset and remains stable under severe class contamination. It recovers transient light-curve classes with high fidelity, while single-source identification is limited by ambiguity in light-curve-derived labels. Our hybrid UQ approach achieves competitive calibration relative to more expensive ensemble baselines. Latent-space analyses indicate that uncertainty aligns with the decision boundary and reveal subclasses within the bogus population. Our results show that injection-driven, weakly supervised training can enable scalable and consistent Real-Bogus classification without human-labeled training data while providing calibrated uncertainties. The method is suited for transfer to forthcoming surveys by re-running the injection-based training pipeline.