artificial intelligence - Looping through training data in Neural Networks Backpropagation Algorithm -

how many times use sample of training data in 1 training cycle? have 60 training data. go through 1st row , forward pass , adjust weights using results backward pass. using sigmoidal function below:

forward pass  si = sum of (wi * uj) ui = f(si) = 1 / 1 + e^ - si  backward pass  output cell = (expected -ui)(f'(si)),  f'(si) = ui(1-ui) 

do go through 2nd row , same process 1st or go around 1st row until error less?

i hope can please

training network

you should use each instance of training set once per training epoch.

a training epoch complete cycle through dataset.

after you've looped through dataset , calculated deltas, should adjust weights of network. may perform new forward pass on neural network , training epoch, looping through training dataset.

graphical representation
great graphical representation of backpropagation may found at link.

---

single-step training

there 2 approaches train network perform classification on dataset. easiest method called single-step or online learning. method find in litterature, , fastest converge. train network calculate deltas each layer , adjust weights each instance of dataset.

thus if have dataset of 60 instances, means should have adjusted weights 60 times before training epoch over.

batch training

the other approach called batch training or offline learning. approach yields network lower residual error. when train network should calculate deltas each layer every instance of dataset, , average individual deltas , correct weights once per epoch.

if have dataset of 60 instances, means should have adjusted weights once before training epoch over.