- 27 view
Neural Net Learning with Most Decent Method part5
- Set initial value to weights and threshold(weight_input_hidden, weight_hidden_output, threshold_hidden), prepare num which is iteration number, δwhich is acceptable difference and learning rate α.
- Obtain below with input data x^p, p=1,・・・,P.
- Obtain gradients with teacher value d^p, p=1,・・・,P.
- Update parameters using gradient.
- When number of iteration reaches num or square error is lower than δ, stop calculation. If it’s not, go back to step2.