18:23:50	 From  Alberto Iess : I agree, more than 10000 points is already too much for LSTM
18:27:05	 From  Wathela Alhassan : I couldn’t find the link to the code, can you please post it here.
18:27:23	 From  Gabriele Vajente (he/him) : https://git.ligo.org/gabriele-vajente/nonsens
18:27:30	 From  Wathela Alhassan : Thank you
18:41:16	 From  Luigia Petre : https://towardsdatascience.com/deeper-neural-networks-lead-to-simpler-embeddings-3c393f5aed04
18:44:15	 From  Bas Swinkels : @Gabriele Barbara Patricella's tool is decribed here: https://tds.virgo-gw.eu/ql/?c=14291
18:45:27	 From  Bas Swinkels : If I understand correctly, they multiply the known calibration line with many aux channels, then look for coherence at the frequency of the line. A faster way, as you said, would be to demodulate hrec with the known line, than look for coherence between the demodulated signal and all aux signals at low frequencies, Bruco style
18:48:11	 From  Alberto Iess : https://www.ams.org/journals/jams/2016-29-04/S0894-0347-2016-00852-4/
18:48:22	 From  Alberto Iess : https://arxiv.org/abs/1608.08225
18:49:58	 From  Didier Verkindt : Fully agree with Gabriele!
18:50:20	 From  Andrei Utina : +1
18:58:45	 From  Marco Cavaglia : bye!