神经网络

Program exercise

• Feedforward and cost function
  • 　　complete the code in nnCostFunction.m
  •       costFunction is
  •  
  • The input Matrix X is a 5000 * 400 matrix， and the output vector y is 5000 * 1.
  • The initial theta provided by assignment is a vetor and need to be urolled into matrix. This procedure can be done by using the function           reshape(X, M, N). This function return a M*N matrix whose element is clockwise of vector X. An error is returned if X does not have M*N element. And, we can use theta1 directly because the code have reshape the matrix for us.
  • when I calculate the transform_y, I made I stupid mistake ...
  • another mistake , forget to remove theta1 when do the regulization cost funtion. And forget how to trim a matrix