cs231n作业：Assignment1-SVM

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

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

                loss += margin

    # 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
    dW += reg*2*W

    # Add regularization to the loss.
    loss += reg * np.sum(W * W)

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
    num_train = X.shape[0]

    results = X.dot(W)

    correct_results = results[np.arange(num_train),y]#一维
    correct_results = np.reshape(correct_results,(num_train,-1))#二维

    results_modify = results - correct_results + 1
    results_modify = np.maximum(0,results_modify)
    results_modify[np.arange(num_train),y] = 0

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

    #向量化计算的骚套路

    results_modify[results_modify>0] = 1#计算贡献次数
    row = np.sum(results_modify,axis=1)#当loss大于0时，正确的类别的梯度需要计算的贡献次数
    #print(row.T.shape)
    results_modify[np.arange(num_train),y] = -row.T#正确的类别
    #print(results_modify[np.arange(num_train),y])

    dW = (X.T).dot(results_modify)
    dW /= num_train
    dW += reg*2*W

    pass

    # *****END OF YOUR CODE (DO NOT DELETE/MODIFY THIS LINE)*****

    return loss, dW


#end