.@physicsscotland @UofGravity #ConvolutionalNeuralNetworks (6 conv & 3 hidden layers) 2 classify simulated #GravitationalWave transients 4m 0-spin #BBH 5M☉, 95M mergers; 1E3 unique wave4ms; 4E5 'whitened' training samples (50% S, 50% S+N) @LIGO @ego_virgo https://t.co/MlIyydQARU pic.twitter.com/4htJi2oxyA— Satyen Baindur (@Satyen_Baindur) December 20, 2017
#CNN #ConvolutionalNeuralNetwork does better than #MatchedFilter at low #SNR & low #FalseAlarmRate in classification task: even w #GaussianNoise, where MF is expected 2 b optimal, which encourages expectation #DeepLearning might do better/faster #GW detection in #NonGaussianNoise— Satyen Baindur (@Satyen_Baindur) December 20, 2017
