吴恩达深度学习L1W2作业1

ipython：

import numpy as np
def sigmoid(x):

    s = 1/(1 + np.exp(-x))

    return s 

x = np.array([1,2,3])
sigmoid(x)

array([0.73105858, 0.88079708, 0.95257413])

创建函数sigmoid_grad（）计算sigmoid函数相对于其输入x的梯度

def sigmoid_derivative(x):
   
    s = sigmoid(x)
    ds = s * (1 - s)

    return ds
x = np.array([1, 2, 3])
print ("sigmoid_derivative(x) = " + str(sigmoid_derivative(x)))

sigmoid_derivative(x) = [0.19661193 0.10499359 0.04517666]

import numpy as np
def image2vector(image):
    v = image.reshape(image.shape[0] * image.shape[1] * image.shape[2], 1)
    print(image.shape[0])#读取矩阵第一维度，3个小矩阵
    print(image.shape[1])#读取矩阵第二维度，小矩阵李有3个小小矩阵
    print(image.shape[2])#读取矩阵第三维度，小小矩阵有两个数据
    #不能单纯的理解shape[0]为矩阵的行数
    print(image.shape[0] * image.shape[1] * image.shape[2])
    return v

image = np.array([[[ 0.67826139, 0.29380381],
          [ 0.90714982, 0.52835647],
          [ 0.4215251 , 0.45017551]],
 
         [[ 0.92814219, 0.96677647],
          [ 0.85304703, 0.52351845],
          [ 0.19981397, 0.27417313]],
 
         [[ 0.60659855, 0.00533165],
          [ 0.10820313, 0.49978937],
          [ 0.34144279, 0.94630077]]])

print(str(image2vector(image)))

3
3
2
18
[[0.67826139]
 [0.29380381]
 [0.90714982]
 [0.52835647]
 [0.4215251 ]
 [0.45017551]
 [0.92814219]
 [0.96677647]
 [0.85304703]
 [0.52351845]
 [0.19981397]
 [0.27417313]
 [0.60659855]
 [0.00533165]
 [0.10820313]
 [0.49978937]
 [0.34144279]
 [0.94630077]]

import numpy as np

def normalizeRows(x):
    x_norm = np.linalg.norm(x, axis = 1, keepdims = True)#axis = 1 行；axis = 0 列；keepdim = True 表示保持其维度
    x = x / x_norm
    print(x_norm.shape)
    print(x.shape)
    return x

x = np.array([
    [0, 3, 4],
    [1, 6, 4]])
print("normalizeRows(x) = " + str(normalizeRows(x)))

(2, 1)
(2, 3)
normalizeRows(x) = [[0.         0.6        0.8       ]
 [0.13736056 0.82416338 0.54944226]]

import numpy as np
def softmax(x):
    x_exp = np.exp(x)
    x_sum = np.sum(x_exp, axis = 1, keepdims = True)
    s = x_exp / x_sum
    print(x_exp)
    print("x_sum:" + str(x_sum))
    return s

x = np.array([
    [9, 2, 5, 0, 0],
    [7, 5, 0, 0 ,0]])
print("softmax(x) = " + str(softmax(x)))

[[8.10308393e+03 7.38905610e+00 1.48413159e+02 1.00000000e+00
  1.00000000e+00]
 [1.09663316e+03 1.48413159e+02 1.00000000e+00 1.00000000e+00
  1.00000000e+00]]
x_sum:[[8260.88614278]
 [1248.04631753]]
softmax(x) = [[9.80897665e-01 8.94462891e-04 1.79657674e-02 1.21052389e-04
  1.21052389e-04]
 [8.78679856e-01 1.18916387e-01 8.01252314e-04 8.01252314e-04
  8.01252314e-04]]

np.dot()函数主要有两个功能，向量点积和矩阵乘法

 np.dot(a, b), 其中a为一维的向量，b为一维的向量，当然这里a和b都是np.ndarray类型的, 此时因为是一维的所以是向量点积

 np.dot(a, b), 其中a为二维矩阵，b为一维向量，这时b会被当做一维矩阵进行计算

 注意的是一维矩阵和一维向量的区别，一维向量的shape是(5, ), 而一维矩阵的shape是(5, 1)

 np.dot(a ,b), 其中a和b都是二维矩阵，此时dot就是进行的矩阵乘法运算
 注意:不同于np.multiply()和* 操作符（相当于Matlab / Octave中的 .*）执行逐元素的乘法，np.dot()执行的是矩阵-矩阵或矩阵向量乘法，

import numpy as np
x1 = [1,2,3]
x2 = [4,5,6]
dot= np.dot(x1,x2)
print(dot)

32

np.outer()函数

①对于多维向量，全部展开变为一维向量

②第一个参数表示倍数，使得第二个向量每次变为几倍。

③第一个参数确定结果的行，第二个参数确定结果的列

import numpy as np
x1 = [1,2,3]
x2 = [4,5,6]
outer = np.outer(x1,x2)
print(outer)
x1 = [[1,2],[3,4]]
x2 = [[1,1],[1,1]]
outer = np.outer(x1,x2)
print(outer)

[[ 4  5  6]
 [ 8 10 12]
 [12 15 18]]
[[1 1 1 1]
 [2 2 2 2]
 [3 3 3 3]
 [4 4 4 4]]

np.multiply()

①multiply(a,b)就是个乘法，如果a,b是两个数组，那么对应元素相乘

②两个参数和结果的shape应该一致,不一致，在计算中会进行广播处理

import numpy as np
x1 = [1,2,3]
x2 = [4,5,6]
mul = np.multiply(x1,x2)
print(mul)

x1 = np.array([[1,2,3]])
x2 = np.array([4,5,6])
mul = np.multiply(x1,x2)
print(x1.shape)
print(x2.shape)
print(mul)

[ 4 10 18]
(1, 3)
(3,)
[[ 4 10 18]]

L1损失函数

def L1(yhat, y):
    #yhat:预测标签 y:实际标签
    loss = np.sum(np.abs(y - yhat)) #abs(x): x的绝对值
    
    return loss

yhat = np.array([.9, 0.2, 0.1, .4, .9])
y = np.array([1, 0, 0, 1, 1])
print("L1 = " + str(L1(yhat,y)))

L1 = 1.1

L2损失函数

def L2(yhat, y):

    loss = np.dot((y - yhat),(y - yhat).T) #  .T 矩阵的转置
    
    return loss

yhat = np.array([.9, 0.2, 0.1, .4, .9])
y = np.array([1, 0, 0, 1, 1])
print("L2 = " + str(L2(yhat,y)))

L2 = 0.43