public List <string> Predict(TensorOld X)
{
X = Normalize(X);
LastRawResult = Model.Predict(X);
LastCodecResult = Utilities.ProbabilityToCode(LastRawResult);
if (LabelCodec != null)
{
return(LabelCodec.Decode(LastCodecResult));
}
return(LastRawResult.GetRawValues().Select(a => a.ToString()).ToList());
}
private void SetTrainingData(TensorOld trainX, TensorOld trainY, TensorOld testX, TensorOld testY)
{
if (Normalizer != null)
{
TrainX = Normalizer.Normalize(trainX);
if (testX != null)
{
TestX = Normalizer.Normalize(testX);
}
}
else
{
TrainX = trainX;
TestX = testX;
}
if (LabelCodec != null)
{
TrainY = LabelCodec.Encode(trainY.GetRawValues());
if (testY != null)
{
TestY = LabelCodec.Encode(testY.GetRawValues());
}
}
else
{
TrainY = trainY;
TestY = testY;
}
var shape = TrainX.Shape;
shape[0] = BatchSize;
xBuff = new TensorOld(shape);
yBuff = new TensorOld(BatchSize, TrainY.shape[1]);
sampleCount = TrainX.shape[0];
xWidth = TrainX.ElementCount / TrainX.shape[0];
yWidth = TrainY.ElementCount / TrainY.shape[0];
batchPerEpoch = trainX.shape[0] / BatchSize;
if (trainX.shape[0] % BatchSize != 0)
{
batchPerEpoch++;
}
}
