Deep Learning without Weight Transport

In a typical deep-learning network, some signals flow along a forward path through multiple layers of processing units from the input layer to the output, while other signals flow back from the output layer along a feedback path . Forward-path signals perform inference (e.g. they try to infer what objects are