Reputation: 33
My question is if we have to use a trained neural network in an algorithm how to determine its timing complexity? or how many multiplications are done to generate the output?
any comments greatly appreciated. Thank in advance.
Upvotes: 3
Views: 559
Reputation: 66
These might help you to decide. NN timing complexity
And for number of multiplication you may find this and this useful.
Since you've saved the NN. you just need to test it or just use the NN in feed forward phase. So your calculations seems to be correct. Note that you have to involve the number of testing samples and applying the activation functions in your calculations as well for better describing the behavior of the NN.
Hope that help.
Upvotes: 3
Reputation: 10151
Complexity is not always measured in number of multiplications.
But if this is what you want to want to take as a metric for your complexity, you first need to look in the implementation of nodes you use in your neural network first of all to perform the action of the neuron and than how many multiplications are used for each training-step.
Only After you know all that you can look for the number of nodes and connections in your network.
Upvotes: 0