All Convolutions in a very dense block are ReLU-activated and use batch normalization. Channel-wise concatenation is just feasible if the peak and width dimensions of the info keep on being unchanged, so convolutions inside of a dense block are all of stride one. Pooling layers are inserted between dense https://financefeeds.com/coins-listing-this-week-best-new-copyright-coins-to-buy-now-for-535-roi-in-february-first-week/