/// <summary>
/// 分类问题中产生的如果是概率，需要把概率转换为编码结果
/// 后面再解码得到最终的分类
/// </summary>
/// <param name="probability">概率</param>
/// <param name="code">编码</param>
public static void ProbabilityToCode(TensorOld probability, TensorOld code)
{
if (probability.Rank == 1)
{
probability = probability.Reshape(1, probability.ElementCount);
}
if (probability.Rank != 2)
{
throw new Exception("to do codec, Rank must be 2");
}
if (probability.shape[1] == 1)
{
TensorOld.Apply(probability, code, a => a > 0.5 ? 1 : 0);
}
else
{
var buff = new double[probability.shape[1]];
for (int i = 0; i < probability.shape[0]; i++)
{
probability.GetByDim1(i, buff);
ComputeCode(buff);
Array.Copy(buff, 0, code.GetRawValues(), i * buff.Length, buff.Length);
}
}
}
private void ComputeCrossEntropy(TensorOld y, TensorOld yHat)
{
var foreoutData = ForwardOutput.GetRawValues();
for (int i = 0; i < sampleNumber; i++)
{
//取出一个样本及其对应的Label
y.GetByDim1(i, yBuff);
yHat.GetByDim1(i, yHatBuff);
//计算交叉熵
foreoutData[i] = Functions.CrossEntropy(yBuff, yHatBuff);
}
Array.Copy(y.values, 0, BackwardOutput.values, 0, y.ElementCount);
}
public override double GetLoss(TensorOld y, TensorOld yHat)
{
var outData = ForwardOutput.GetRawValues();
var result = 0d;
for (int i = 0; i < y.shape[0]; i++)
{
//取出一个样本及其对应的Label
y.GetByDim1(i, yBuff);
yHat.GetByDim1(i, yHatBuff);
//计算交叉熵
result += Functions.CrossEntropy(yBuff, yHatBuff);
}
return(result / sampleNumber);
}
private void ComputeCrossEntropy(TensorOld y, TensorOld yHat)
{
var forwardoutData = ForwardOutput.GetRawValues();
var backwardoutData = BackwardOutput.GetRawValues();
for (int i = 0; i < sampleNumber; i++)
{
//取出一个样本及其对应的Label
y.GetByDim1(i, yBuff);
yHat.GetByDim1(i, yHatBuff);
//计算交叉熵
forwardoutData[i] = Functions.CrossEntropy(yBuff, yHatBuff);
//计算损失函数关于输入的导数
Derivatives.CrossEntropy(yBuff, yHatBuff, derBuff);
Array.Copy(derBuff, 0, backwardoutData, i * derBuff.Length, derBuff.Length);
}
}
