Search code examples
javadeeplearning4j

I need to understand what it is nEpoch in sample

I am running the example number

[96] org.deeplearning4j.examples.convolution.objectdetection.HouseNumberDetection.

It's in the training phase and I was wondering why this sentence keeps coming up:

o.d.o.l.ScoreIterationListener - Score at iteration 352 is 2.6203769938063846

Specifically, I do not understand the number 352, in this case, if by invoking the sentence that causes the above, that is, this

model.fit(train, nEpochs);

nEpochs is instantiated as

int nEpochs = 20;

From ignorance and wanting to understand, shouldn't there be at most 20 scores? 20 iterations to train and after that, save the model?

I am using latest version

    <version>1.0.0-beta6</version>

Thank you.

Solution

  • Iterations aren't the same thing as epochs. An epoch is when all your data has been used once. An iteration, in DL4J is when a mini batch has been used. So if you have 100 examples in you dataset and you use 10 each in your mini batches, then you will have 10 iterations per epoch.