All Convolutions in a dense block are ReLU-activated and use batch normalization. Channel-wise concatenation is only feasible if the peak and width dimensions of the data continue to be unchanged, so convolutions in the dense block are all of stride 1. Pooling layers are inserted between dense blocks for https://financefeeds.com/the-rise-of-copy-trading-in-mena-key-drivers-and-insights/