svm

import numpy as np

from random import shuffle

def svm_loss_naive(W, X, y, reg):

"""

Structured SVM loss function, naive implementation (with loops).

Inputs have dimension D, there are C classes, and we operate on minibatches

of N examples.

Inputs:

- W: A numpy array of shape (D, C) containing weights.

- X: A numpy array of shape (N, D) containing a minibatch of data.

- y: A numpy array of shape (N,) containing training labels; y[i] = c means

that X[i] has label c, where 0 <= c < C.

- reg: (float) regularization strength

Returns a tuple of:

- loss as single float

- gradient with respect to weights W; an array of same shape as W

"""

dW = np.zeros(W.shape) # initialize the gradient as zero

# compute the loss and the gradient

num_classes = W.shape[1]

num_train = X.shape[0]

loss = 0.0

for i in xrange(num_train):

scores = X[i].dot(W)#X[i]=[2,5,4,8,9,9,4,5,5], W=D*C, Scores=[c1 c2 c3 c4 c5 c6 ...cc]

correct_class_score = scores[y[i]]#correct_class_score=  >  scores[y[i]]=>scores[y[1]]=scores[6]=c6

for j in xrange(num_classes):#num_classes=C  j=1,2,3,4,5,6,...,c

if j == y[i]:#if 1 == y[1]=6, go------

continue

margin = scores[j] - correct_class_score + 1 # note delta = 1

if margin > 0:

loss += margin

dW[:,j] += X[i].T

dW[:,y[i]] += -X[i].T

# Right now the loss is a sum over all training examples, but we want it

# to be an average instead so we divide by num_train.

loss /= num_train

dW /= num_train

# Add regularization to the loss.

loss += 0.5 * reg * np.sum(W * W)

dW += reg * W

#############################################################################

# TODO:                                                                    #

# Compute the gradient of the loss function and store it dW.                #

# Rather that first computing the loss and then computing the derivative,  #

# it may be simpler to compute the derivative at the same time that the    #

# loss is being computed. As a result you may need to modify some of the    #

# code above to compute the gradient.                                      #

#############################################################################

return loss, dW

def svm_loss_vectorized(W, X, y, reg):

"""

Structured SVM loss function, vectorized implementation.

Inputs and outputs are the same as svm_loss_naive.

"""

loss = 0.0

dW = np.zeros(W.shape) # initialize the gradient as zero

#############################################################################

# TODO:                                                                    #

# Implement a vectorized version of the structured SVM loss, storing the    #

# result in loss.                                                          #

#############################################################################

num_train = X.shape[0]

num_classes = W.shape[1]

scores = X.dot(W)

correct_class_scores = scores[range(num_train), list(y)].reshape(-1,1) #(N, 1)

margins = np.maximum(0, scores - correct_class_scores +1)

margins[range(num_train), list(y)] = 0

loss = np.sum(margins) / num_train + 0.5 * reg * np.sum(W * W)

#pass

#############################################################################

#                            END OF YOUR CODE                              #

#############################################################################

#############################################################################

# TODO:                                                                    #

# Implement a vectorized version of the gradient for the structured SVM    #

# loss, storing the result in dW.                                          #

#                                                                          #

# Hint: Instead of computing the gradient from scratch, it may be easier    #

# to reuse some of the intermediate values that you used to compute the    #

# loss.                                                                    #

#############################################################################

coeff_mat = np.zeros((num_train, num_classes))

coeff_mat[margins > 0] = 1

coeff_mat[range(num_train), list(y)] = 0

coeff_mat[range(num_train), list(y)] = -np.sum(coeff_mat, axis=1)

dW = (X.T).dot(coeff_mat)

dW = dW/num_train + reg*W

#pass

#############################################################################

#                            END OF YOUR CODE                              #

#############################################################################

return loss, dW

你可能感兴趣的:(svm)