Optimized convolutional neural networks for the detection of multimodal gravitational wave signals

Gravitational wave astronomy can benefit from the rapid classification of gravitational wave signals buried deep in instrumentation noise. In 2017, George and Huerta (and since then other researchers) have considered Convolutional Neural Networks to detect gravitational wave signals and estimate some of the corresponding binary's parameters. In this talk, I will describe extensions of this classification strategy. In particular, we discuss strategies to optimize the hyperparameters of our network, in an attempt to make our networks as compact and effective as possible. Results will be discussed for training data using models with and without subdominant modes.