artificial intelligence - How does a back-propagation training algorithm work? -
artificial intelligence - How does a back-propagation training algorithm work? -
i've been trying larn how propagation works neural networks yet find explanation less technical aspect (i have searched places , found http://www.alexandria.nu/ai/neural_net_demo/ useful far). how propagation work? how larn training dataset provided? have code until then, need gain stronger understanding of them.
back-propagation works in logic similar of feed-forward. difference direction of info flow. in feed-forward step, have inputs , output observed it. can propagate values forwards train neurons ahead.
in back-propagation step, cannot know errors occurred in every neuron ones in output layer. calculating errors of output nodes straightforward - can take difference between output neuron , desired output actual output instance in training set. neurons in hidden layers must update errors this. have pass error values them. these values, hidden neurons can update error , other parameters using weighted sum of errors layer ahead.
a step-by-step demo of feed-forward , back-propagation steps can found here.
edit:
if you're beginner neural networks, can begin learning perceptron, advance nn, multilayer perceptron.
artificial-intelligence computer-science neural-network backpropagation
Comments
Post a Comment