This article provides two valuable services to the reader: first, it offers a method to replace the many bad network diagrams that are out there with good ones, and second, it describes each of the components of those diagrams - and therefore or neural networks - clearly and in detail. Things like weights, bias and activation functions are key to describing different types of network, and these new diagrams make all of these clear (and even describe the matrices on which the mathematical operations actually occur).
Today: 0 Total: 1120 [Direct link] [Share]
