CS231n-assignment1 SVM 的梯度计算(dW) 用FOR循环实现和用矩阵实现

 

 

1. 

1. FOR循环实现

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 range(num_train):
    scores = X[i].dot(W)
    correct_class_score = scores[y[i]]
    for j in range(num_classes):
      if j == y[i]:
        continue
      margin = scores[j] - correct_class_score + 1 # note delta = 1
      if margin > 0:
        loss += margin
        dW[:,y[i]] = dW[:,y[i]] - X[i,:] 
        dW[:,j] = dW[:,j] + X[i,:] 

    #dW = dW+dWi    
  # 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 += 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_classes = W.shape[1]
  num_train = X.shape[0]
  loss = 0.0  
  scores = X.dot(W)                    # N by C
  #select element in row i, col y[i] as correct_class_score  
  #print(scores.shape)
  correct_class_score = scores[np.arange(num_train), y]             #N by 1
  #print(correct_class_score.shape)
  correct_class_score = np.reshape(correct_class_score, (num_train, 1)) 
  #print(correct_class_score.shape)
  margin = np.maximum(0,scores - correct_class_score + 1)
    
  #set the margin in j = yi to 0
  margin[np.arange(num_train), y] = 0.0
  #set the margin <= 0  to 0
  margin[margin <= 0] = 0.0
  loss += np.sum(margin)    
  loss /= num_train
  loss += reg * np.sum(W * W)
  #############################################################################
  #                             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.                                                                     #
  #############################################################################
  #sc = np.maximum(0,scores - correct_class_score + 1)
  #dW = np.dot(X.T ,np.asarray(sc>0,'i'))/num_train + reg * W 

  margin[margin > 0] = 1.0                   # N by C
  row_sum = np.sum(margin, axis=1)              # 1 by N
  margin[np.arange(num_train), y] = -row_sum        
  dW += np.dot(X.T, margin)/num_train + reg * W     # D by C
  #############################################################################
  #                             END OF YOUR CODE                              #
  #############################################################################

  return loss, dW

2. 矩阵实现

 

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_classes = W.shape[1]
  num_train = X.shape[0]
  loss = 0.0  
  scores = X.dot(W)
  #select element in row i, col y[i] as correct_class_score  
  correct_class_score = scores[np.arange(num_train), y]
  correct_class_score = np.reshape(correct_class_score, (num_train, 1)) 
  margin = np.maximum(0,scores - correct_class_score + 1)
  #set the margin in j = yi to 0
  margin[np.arange(num_train), y] = 0.0
  #set the margin <= 0  to 0
  margin[margin <= 0] = 0.0
  loss += np.sum(margin)    
  loss /= num_train
  loss += reg * np.sum(W * W)
  #############################################################################
  #                             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.                                                                     #
  #############################################################################
  pass
  #############################################################################
  #                             END OF YOUR CODE                              #
  #############################################################################

  return loss, dW