Learning Without Feedback: Fixed Random Learning Signals Allow for Feedforward Training of Deep Neural Networks
MyNeuroNews,
While the backpropagation of error algorithm enables deep neural network training, it implies (i) bidirectional synaptic weight…