public Result Compress(int[,] imgArray)
{
int totalFrames = (imgArray.GetLength(0) * imgArray.GetLength(1)) / (FrameSize * FrameSize);
int numOfFramesInrow = imgArray.GetLength(0) / FrameSize;
int counter = 0;
Result result = new Result(totalFrames);
result.FrameSize = FrameSize;
for (int i = 0; i < numOfFramesInrow; i++)
{
for (int j = 0; j < numOfFramesInrow; j++)
{
int[] frameBeginning = { FrameSize *i, FrameSize *j };
int[] frame = FrameHelper.GetFrame(frameBeginning, imgArray, FrameSize);
double squareOfVectorSum = FrameHelper.CountSquareOfVectorSum(frame);
double[] normalizedFrame = FrameHelper.Normalize(frame, squareOfVectorSum);
int bestNeuronIndex = FindBestNeuron(normalizedFrame);
result.NeuronIndexes[counter] = bestNeuronIndex;
result.Lenghts[counter] = squareOfVectorSum;
counter++;
}
}
if (counter != totalFrames)
{
throw new Exception("Counter must be equal to total frames");
}
for (int i = 0; i < Neurons.Length; i++)
{
result.IndexToWeights.Add(i, Neurons[i].Weights);
}
return(result);
}
public void Train(int numberOfEpochs, int[,] imgArray)
{
for (int i = 0; i < numberOfEpochs; i++)
{
int[] frameBeginning = FrameHelper.SelectRandomFrameBeginning(imgArray.GetLength(0), FrameSize, random);
int[] frame = FrameHelper.GetFrame(frameBeginning, imgArray, FrameSize);
double[] normalizedFrame = FrameHelper.Normalize(frame, FrameHelper.CountSquareOfVectorSum(frame));
Train(normalizedFrame);
}
}
private double[] NormalizeWeights(double[] weights)
{
double[] normalizedWeights = new double[weights.Length];
double squareOfVectorSum = FrameHelper.CountSquareOfVectorSum(weights);
for (int i = 0; i < weights.Length; i++)
{
if (squareOfVectorSum == 0)
{
normalizedWeights[i] = 0;
}
else
{
normalizedWeights[i] = weights[i] / squareOfVectorSum;
}
}
return(normalizedWeights);
}
