A contrastive self-supervised convolutional autoencoder detects core-collapse supernova gravitational waves with performance comparable to supervised CNNs, better generalization to unseen waveforms, and ~120 kpc sensitive distance under Einstein Telescope noise.
Determine the Core Structure and Nuclear Equation of State of Rotating Core-Collapse Supernovae with Gravitational Waves by Convolutional Neural Networks
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
abstract
Detecting gravitational waves from a nearby core-collapse supernova would place meaningful constraints on the supernova engine and nuclear equation of state. Here we use Convolutional Neural Network models to identify the core rotational rates, rotation length scales, and the nuclear equation of state (EoS), using the 1824 waveforms from Richers et al. (2017) for a 12 solar mass progenitor. High prediction accuracy for the classifications of the rotation length scales ($93\%$) and the rotational rates ($95\%$) can be achieved using the gravitational wave signals from -10 ms to 6 ms core bounce. By including additional 48 ms signals during the prompt convection phase, we could achieve $96\%$ accuracy on the classification of four major EoS groups. Combining three models above, we could correctly predict the core rotational rates, rotation length scales, and the EoS at the same time with more than $85\%$ accuracy. Finally, applying a transfer learning method for additional 74 waveforms from FLASH simulations (Pan et al. 2018), we show that our model using Richers' waveforms could successfully predict the rotational rates from Pan's waveforms even for a continuous value with a mean absolute errors of 0.32 rad s$^{-1}$ only. These results demonstrate a much broader parameter regimes our model can be applied for the identification of core-collapse supernova events through GW signals.
citation-role summary
background 1
citation-polarity summary
years
2026 3roles
background 1polarities
background 1representative citing papers
LightGBM and other regressors achieve R^{2}≈0.6–0.7 under random CV on CCSN GW catalogues but collapse to worse-than-mean performance under Leave-One-EoS-Out validation, exposing a generalisation gap for unseen EoS families.
citing papers explorer
-
Contrastive self-supervised convolutional autoencoder for core-collapse supernova gravitational-wave detection
A contrastive self-supervised convolutional autoencoder detects core-collapse supernova gravitational waves with performance comparable to supervised CNNs, better generalization to unseen waveforms, and ~120 kpc sensitive distance under Einstein Telescope noise.
-
The Generalization Gap in Machine Learning EoS Inference from Core-Collapse Supernova Gravitational Waves
LightGBM and other regressors achieve R^{2}≈0.6–0.7 under random CV on CCSN GW catalogues but collapse to worse-than-mean performance under Leave-One-EoS-Out validation, exposing a generalisation gap for unseen EoS families.
- Parameter Estimation Horizon of Core-Collapse Supernovae with Current and Next-Generation Gravitational-Wave Detectors