From line 149 onwards there doesn't seem to be any mention of what SNR the training injections were made with. It just says 'predefined optimal SNR' but not the actual value(s). Later in the paper there seem to be various 'experiments' with rho_opt = 2,4,6,8,10,12 - does this mean that the NN was trained differently for each experiment with that specific rho_opt value? This is not clear.
Fergus: I feel this is not explained badly in the paper as it is. Maybe though we should include some mention of the varying SNR datasets we used when in the 'simulation details' section.
From line 149 onwards there doesn't seem to be any mention of what SNR the training injections were made with. It just says 'predefined optimal SNR' but not the actual value(s). Later in the paper there seem to be various 'experiments' with rho_opt = 2,4,6,8,10,12 - does this mean that the NN was trained differently for each experiment with that specific rho_opt value? This is not clear.
Fergus: I feel this is not explained badly in the paper as it is. Maybe though we should include some mention of the varying SNR datasets we used when in the 'simulation details' section.