Basic DeepBeliefNetwork example?

[varrav]

Could someone post a simple Accord .NET deep learning example? I've seen the sample one included with the source code, but I'm looking for a bare-minimum example. Something like the following, how would you perform deep learning on it? Thanks.

double inputs[][];
int outputs[]; // 0, 1

// Initialize network.
DeepBeliefNetwork network = new DeepBeliefNetwork(new BernoulliFunction(), 1024, 50, 2);
new GaussianWeights(network).Randomize();
network.UpdateVisibleWeights();

// Train
???

// Test result
for (int i=0; i<inputs.Length; i++)
{
   int testOutput = network.Compute(inputs[i]);
   if (testOutput == outputs[i])
   {
       // Correct!
   }
}

[varrav]

After digging into the demo a bit, I put together a simple deep-learning example for XOR. If anyone else is looking for a basic example, the code is available on github.

The branch "XOR" contains the code sample.

The master branch contains the ASCII digits code sample.

Btw, perhaps this could be useful to others, as an example, in the Accord .NET class documentation for DeepBeliefNetwork?

[César]

Hello Varrav,

Many thanks for the sample! It is definitely going to be added to the documentation. I was planning to create a documentation demo as well but was still trying to find enough time to do it properly, thanks for the amazing help!

Best regards,

Cesar