This article describes work showing how back propagation might work in a human brain. In back propagation, a neural network learns (that is, a neural network adjusts connection weights) after processing some input as a result of signals sent back against the flow of processing. In a human this seemed impractical because it would require the senses to shut down. The work describes here shows that neuron 'bursts' could perform the back propagation function. "Naud and Richards' team proposed that neurons have separate compartments at their top and bottom that process the neural code in completely different ways." One would interpret a burst as a corrective signal, while the other would interpret it as a normal signal.
Today: 3 Total: 89 [Direct link] [Share]
