Learning and interpreting gravitational-wave features from CNNs with a random forest approach
JournalMachine Learning: Science and Technology 6(3), 035045 (2025)
First public: Published online:
- Detection & denoising
- AI for science
Research summary
A hybrid convolutional-neural-network and random-forest pipeline examines how learned features contribute to gravitational-wave detection. Physically motivated quantities derived from feature maps complement the network output in classification. Experiments evaluate detection performance and feature attribution, connecting domain-informed measurements with more interpretable decisions on noisy strain data.
Extended notes
The classifier combines CNN outputs with quantities related to feature-map variance, signal strength, overlap, and amplitude. Random-forest attribution is used to examine their contributions to detection decisions.