使用logloss和L2正则化计算SGD分类器而不使用sklearn，我无法计算损失和损失计算中的错误

Question

def logloss(y_true,y_pred):                                     # compute log-loss                      
    log_loss = (-y_true * math.log(y_pred, 10) - (1 - y_true) * math.log(1 - y_pred,10)).mean()
    return log_loss

def grader_logloss(true,pred):                 # comparing log-loss using assert
    loss = logloss(true,pred)
    assert(loss == 0.07644900402910389)
    return True
true = [1,1,0,1,0]
pred = [0.9,0.8,0.1,0.8,0.2]
grader_logloss(true,pred)

我犯了一个错误

 ---------------------------------------------------------------------------
    TypeError                                 Traceback (most recent call last)
    <ipython-input-110-edd7da954047> in <module>
          5 true = [1,1,0,1,0]
          6 pred = [0.9,0.8,0.1,0.8,0.2]
    ----> 7 grader_logloss(true,pred)

    <ipython-input-110-edd7da954047> in grader_logloss(true, pred)
          1 def grader_logloss(true,pred):
    ----> 2     loss = logloss(true,pred)
          3     assert(loss == 0.07644900402910389)
          4     return True
          5 true = [1,1,0,1,0]

    <ipython-input-109-b96b3bba92ed> in logloss(y_true, y_pred)
          2     '''In this function, we will compute log loss '''
          3     n = len(y_true)
    ----> 4     log_loss = (-y_true * math.log(y_pred, 10) - (1 - y_true) * math.log(1 - y_pred,10)).mean()
          5     return log_loss

    TypeError: bad operand type for unary -: 'list'

我无法获得什么是操作数类型。我已经找过了，但没有弄清楚

预期结果是

True

计算梯度

def gradient_dw(x,y,w,b,alpha,N):
    '''In this function, we will compute the gardient w.r.to w '''
    dw = x*(y - sigmoid(np.dot(w.T,x) + b)) - ((alpha*x)/N)
    return dw

梯度的计算与比较

def grader_dw(x,y,w,b,alpha,N):
    grad_dw=gradient_dw(x,y,w,b,alpha,N)
    assert(grad_dw==2.613689585)
    return True
grad_x=np.array([-2.07864835,  3.31604252, -0.79104357, -3.87045546, -1.14783286,
       -2.81434437, -0.86771071, -0.04073287,  0.84827878,  1.99451725,
        3.67152472,  0.01451875,  2.01062888,  0.07373904, -5.54586092])
grad_y=0
grad_w,grad_b = initialize_weights(grad_x)
alpha=0.0001
N=len(X_train)
grader_dw(grad_x,grad_y,grad_w,grad_b,alpha,N)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-152-b22fd39ec68a> in <module>
     10 alpha=0.0001
     11 N=len(X_train)
---> 12 grader_dw(grad_x,grad_y,grad_w,grad_b,alpha,N)

<ipython-input-152-b22fd39ec68a> in grader_dw(x, y, w, b, alpha, N)
      1 def grader_dw(x,y,w,b,alpha,N):
      2     grad_dw=gradient_dw(x,y,w,b,alpha,N)
----> 3     assert(grad_dw==2.613689585)
      4     return True
      5 grad_x=np.array([-2.07864835,  3.31604252, -0.79104357, -3.87045546, -1.14783286,

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

我不知道我的梯度求值哪里不正确，也不知道为什么即使我尝试了a.any（）或a.all（），断言函数仍然失败

Answer 1

使用numpy.array存储true和pred，并稍微调整代码 使用^{中定义的操作。由于python中的普通list不支持-和*和+等元素操作，因此它没有mean方法

import numpy as np
def logloss(y_true,y_pred):                                     # compute log-loss                      
    log_loss = (-y_true * np.log10(y_pred) - (1 - y_true) * np.log10(1 - y_pred)).mean()
    return log_loss
def grader_logloss(true,pred):                 # comparing log-loss using assert
    loss = logloss(true,pred)
    assert(loss == 0.07644900402910389)
    return True
true = np.array([1,1,0,1,0])
pred = np.array([0.9,0.8,0.1,0.8,0.2])
grader_logloss(true,pred)

新问题的更新： 我假设你的dw是一个标量，因为你有grad_dw==2.613689585。然后，在gradient_dw函数中，更改此行：

dw = x*(y - sigmoid(np.dot(w.T,x) + b)) - ((alpha*x)/N)

这条线

dw = (x*(y - sigmoid(np.dot(w.T,x) + b)) - ((alpha*x))).sum() / N

此外，您应该使用assert grad_dw==2.613689585而不是assert (grad_dw==2.613689585)