private static void GetClassifySampleInputData(ref IEnumerable<BatchInputWrapper> trainData, ref IEnumerable<BatchInputWrapper> cvData)
{
List<BatchInputWrapper> trainDataList = new List<BatchInputWrapper>();
List<BatchInputWrapper> cvDataList = new List<BatchInputWrapper>();
BatchInputWrapper currentBatch = new BatchInputWrapper(5, 1);
currentBatch.AddInputAndTrueLabelData(new float[] { 1, 0, 1, 0, 1 }, new float[] { 1 });
//currentBatch.AddInputAndTrueLabelData(new float[] { 1, 2, 3, 2, 1 }, new float[] { 0 });
currentBatch.SealBatch();
trainDataList.Add(currentBatch);
//currentBatch = new BatchInputWrapper(5, 1);
//currentBatch.AddInputAndTrueLabelData(new float[] { 1, 2, 0, 2, 1 }, new float[] { 0 });
//currentBatch.AddInputAndTrueLabelData(new float[] { 1, 0, 3, 0, 1 }, new float[] { 1 });
//currentBatch.SealBatch();
//trainDataList.Add(currentBatch);
trainData = trainDataList.AsEnumerable();
currentBatch = new BatchInputWrapper(5, 1);
currentBatch.AddInputAndTrueLabelData(new float[] { 1, 0, 1, 1, 1 }, new float[] { 0 });
currentBatch.SealBatch();
cvDataList.Add(currentBatch);
cvData = cvDataList.AsEnumerable();
}
private static void GetClassifySampleInputData(ref IEnumerable <BatchInputWrapper> trainData, ref IEnumerable <BatchInputWrapper> cvData)
{
List <BatchInputWrapper> trainDataList = new List <BatchInputWrapper>();
List <BatchInputWrapper> cvDataList = new List <BatchInputWrapper>();
BatchInputWrapper currentBatch = new BatchInputWrapper(5, 1);
currentBatch.AddInputAndTrueLabelData(new float[] { 1, 0, 1, 0, 1 }, new float[] { 1 });
//currentBatch.AddInputAndTrueLabelData(new float[] { 1, 2, 3, 2, 1 }, new float[] { 0 });
currentBatch.SealBatch();
trainDataList.Add(currentBatch);
//currentBatch = new BatchInputWrapper(5, 1);
//currentBatch.AddInputAndTrueLabelData(new float[] { 1, 2, 0, 2, 1 }, new float[] { 0 });
//currentBatch.AddInputAndTrueLabelData(new float[] { 1, 0, 3, 0, 1 }, new float[] { 1 });
//currentBatch.SealBatch();
//trainDataList.Add(currentBatch);
trainData = trainDataList.AsEnumerable();
currentBatch = new BatchInputWrapper(5, 1);
currentBatch.AddInputAndTrueLabelData(new float[] { 1, 0, 1, 1, 1 }, new float[] { 0 });
currentBatch.SealBatch();
cvDataList.Add(currentBatch);
cvData = cvDataList.AsEnumerable();
}
private static void GetMNISTInputData(ref IEnumerable <BatchInputWrapper> trainData, ref IEnumerable <BatchInputWrapper> cvData)
{
const string filename = @"D:\vinods\Data\MNIST\train.csv";
// label, pixels
List <Tuple <float[], float[]> > inputData = new List <Tuple <float[], float[]> >();
using (StreamReader sr = new StreamReader(filename))
{
sr.ReadLine();
string line;
while ((line = sr.ReadLine()) != null)
{
var values = line.Split(',');
List <float> label = new List <float>();
List <float> pixels = new List <float>();
for (int i = 0; i < values.Count(); i++)
{
if (i == 0)
{
var tmp = Enumerable.Repeat <float>(0, 10);
label.AddRange(tmp);
label[int.Parse(values.ElementAt(i))] = 1;
}
else
{
pixels.Add(float.Parse(values.ElementAt(i)) / (float)255);
}
}
inputData.Add(Tuple.Create <float[], float[]>(label.ToArray(), pixels.ToArray()));
}
}
Random rnd = new Random();
for (int i = inputData.Count - 1; i >= 0; i--)
{
int swapIndex = rnd.Next(i + 1);
var tmp = inputData.ElementAt(i);
inputData[i] = inputData.ElementAt(swapIndex);
inputData[swapIndex] = tmp;
}
int cutIndex = (int)(inputData.Count * 0.8);
int inputDimension = 784, labelDimension = 10;
// Creating train batches
BatchInputWrapper currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
List <BatchInputWrapper> trainDataList = new List <BatchInputWrapper>();
foreach (var item in inputData.GetRange(0, cutIndex))
{
if (!currentBatch.IsFull)
{
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
else
{
currentBatch.SealBatch();
trainDataList.Add(currentBatch);
currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
}
currentBatch.SealBatch();
trainDataList.Add(currentBatch);
trainData = trainDataList.AsEnumerable();
// Creating test batches
currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
List <BatchInputWrapper> testDataList = new List <BatchInputWrapper>();
foreach (var item in inputData.GetRange(cutIndex, inputData.Count - cutIndex))
{
if (!currentBatch.IsFull)
{
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
else
{
currentBatch.SealBatch();
testDataList.Add(currentBatch);
currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
}
currentBatch.SealBatch();
testDataList.Add(currentBatch);
cvData = testDataList.AsEnumerable();
}
private static void GetMNISTInputData(ref IEnumerable<BatchInputWrapper> trainData, ref IEnumerable<BatchInputWrapper> cvData)
{
const string filename = @"D:\vinods\Data\MNIST\train.csv";
// label, pixels
List<Tuple<float[], float[]>> inputData = new List<Tuple<float[], float[]>>();
using (StreamReader sr = new StreamReader(filename))
{
sr.ReadLine();
string line;
while ((line = sr.ReadLine()) != null)
{
var values = line.Split(',');
List<float> label = new List<float>();
List<float> pixels = new List<float>();
for (int i = 0; i < values.Count(); i++)
{
if (i == 0)
{
var tmp = Enumerable.Repeat<float>(0, 10);
label.AddRange(tmp);
label[int.Parse(values.ElementAt(i))] = 1;
}
else
{
pixels.Add(float.Parse(values.ElementAt(i)) / (float)255);
}
}
inputData.Add(Tuple.Create<float[], float[]>(label.ToArray(), pixels.ToArray()));
}
}
Random rnd = new Random();
for (int i = inputData.Count - 1; i >= 0; i--)
{
int swapIndex = rnd.Next(i + 1);
var tmp = inputData.ElementAt(i);
inputData[i] = inputData.ElementAt(swapIndex);
inputData[swapIndex] = tmp;
}
int cutIndex = (int)(inputData.Count * 0.8);
int inputDimension = 784, labelDimension = 10;
// Creating train batches
BatchInputWrapper currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
List<BatchInputWrapper> trainDataList = new List<BatchInputWrapper>();
foreach (var item in inputData.GetRange(0, cutIndex))
{
if (!currentBatch.IsFull)
{
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
else
{
currentBatch.SealBatch();
trainDataList.Add(currentBatch);
currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
}
currentBatch.SealBatch();
trainDataList.Add(currentBatch);
trainData = trainDataList.AsEnumerable();
// Creating test batches
currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
List<BatchInputWrapper> testDataList = new List<BatchInputWrapper>();
foreach (var item in inputData.GetRange(cutIndex, inputData.Count - cutIndex))
{
if (!currentBatch.IsFull)
{
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
else
{
currentBatch.SealBatch();
testDataList.Add(currentBatch);
currentBatch = new BatchInputWrapper(inputDimension, labelDimension);
currentBatch.AddInputAndTrueLabelData(item.Item2, item.Item1);
}
}
currentBatch.SealBatch();
testDataList.Add(currentBatch);
cvData = testDataList.AsEnumerable();
}
