All Convolutions inside of a dense block are ReLU-activated and use batch normalization. Channel-wise concatenation is only probable if the peak and width dimensions of the info stay unchanged, so convolutions inside a dense block are all of stride one. Pooling levels are inserted between dense blocks for further https://financefeeds.com/secs-complaint-against-arete-wealth-may-expose-regulators-own-shortcomings/