public static Set[] CreateSampleSets(
TestNet consumer,
int batchSize,
int totalSets)
{
var sets = new List <Set>();
var builder = BuilderInstance <double> .Volume;
for (var s = 0; s < totalSets; s += batchSize)
{
var batchInputs = consumer
.InputShape
.Select(inputShape =>
{
var inputBatch = Shape.From(inputShape.Dimensions[0], inputShape.Dimensions[1], inputShape.Dimensions[2], batchSize);
return(builder.Random(inputBatch));
}).ToArray();
var outputShape = Shape.From(consumer.OutputShape.Dimensions[0], consumer.OutputShape.Dimensions[1], consumer.OutputShape.Dimensions[2], batchSize);
var tempBatchOutputs = builder.Random(outputShape);
var batchOutputs = builder.SameAs(outputShape);
tempBatchOutputs.Softmax(batchOutputs);
sets.Add(new Set
{
Inputs = batchInputs,
Outputs = batchOutputs
});
}
return(sets.ToArray());
}
private static void ExecuteNeuralNet(
string name,
TestNet net,
int batchSize,
int totalSets,
int iterations)
{
var inputs = CreateSampleSets(net, batchSize, totalSets);
var stopWatch = new Stopwatch();
Console.WriteLine($"- {name} ------");
stopWatch.Restart();
var trainer = new SgdTrainer(net);
trainer.LearningRate = 0.01;
trainer.Momentum = 0.5;
trainer.BatchSize = batchSize;
for (var i = 0; i < iterations; i++)
{
foreach (var set in inputs)
{
trainer.Train(set.Inputs[0], set.Outputs);
}
}
stopWatch.Stop();
Console.WriteLine(" total: {0:0.000}ms", stopWatch.ElapsedMilliseconds);
Console.WriteLine(" forward: {0:0.000}ms", trainer.ForwardTimeMs);
Console.WriteLine(" backward: {0:0.000}ms", trainer.BackwardTimeMs);
Console.WriteLine(" update: {0:0.000}ms", trainer.UpdateWeightsTimeMs);
}
private static TestNet Create(int layerSize, int nmLayers, Shape input, int output)
{
var net = new TestNet();
net.InputShape = new[] { Shape.From(input) };
net.OutputShape = Shape.From(1, 1, output);
net.AddLayer(new InputLayer(input.Dimensions[0], input.Dimensions[1], input.Dimensions[2]));
for (var i = 0; i < nmLayers; i++)
{
net.AddLayer(new FullyConnLayer(layerSize));
net.AddLayer(new ReluLayer());
}
net.AddLayer(new FullyConnLayer(output));
net.AddLayer(new SoftmaxLayer(output));
return(net);
}