Convergence of BP Algorithm with Variable Learning Rates for FNN Training

Convergence results are presented for the batch backpropagation algorithm with variable learning rates for training feedforward neural networks with a hidden layer. The monotonicity of the error function in the training iteration is also proved.