6.034 Artificial Intelligence - Recitations, fall 2004 online slides on learning
	Networks That Deal With Time In the feedforward networks we have been studying, transfer functions capture the network state (not usually spatiotemporal in nature). Recurrent neural networks feed signal back from output to network (output to hidden, hidden to hidden, hidden to input, others). In this way they can learn a function of time. y(t) = w1 x1(t) + w2 x2(t) + ... + wn+1 x1(t-1) + wn+1 x2(t-1) + ...