All Convolutions inside a dense block are ReLU-activated and use batch normalization. Channel-wise concatenation is only feasible if the height and width dimensions of the info continue to be unchanged, so convolutions in a very dense block are all of stride one. Pooling layers are inserted between dense blocks https://financefeeds.com/weekly-roundup-biden-to-pardon-sam-bankman-fried-spotware-unveils-ctrader-5-0/