【发布时间】：2020-05-24 17:57:19
【问题描述】：
我是机器学习的新手，我试图使用 numpy 从头开始实现向量化线性回归。我尝试使用 y=x 测试实现。但是我的损失在增加，我无法理解为什么。如果有人能指出为什么会发生这种情况，那就太好了。提前致谢！
import numpy as np
class LinearRegressor(object):
def __init__(self, num_features):
self.num_features = num_features
self.w = np.random.randn(num_features, 1).astype(np.float32)
self.b = np.array(0.0).astype(np.float32)
def forward(self, x):
return np.dot(x, self.w) + self.b
@staticmethod
def loss(y_pred, y_true):
l = np.average(np.power(y_pred - y_true, 2)) / 2
return l
def calculate_gradients(self, x, y_pred, y_true):
self.dl_dw = np.dot(x.T, y_pred - y_true) / len(x)
self.dl_db = np.mean(y_pred - y_true)
def optimize(self, step_size):
self.w -= step_size*self.dl_dw
self.b -= step_size*self.dl_db
def train(self, x, y, step_size=1.0):
y_pred = self.forward(x)
l = self.loss(y_pred=y_pred, y_true=y)
self.calculate_gradients(x=x, y_pred=y_pred, y_true=y)
self.optimize(step_size=step_size)
return l
def evaluate(self, x, y):
return self.loss(self.forward(x), y_true)
check_reg = LinearRegressor(num_features=1)
x = np.array(list(range(1000))).reshape(-1, 1)
y = x
losses = []
for iteration in range(100):
loss = check_reg.train(x=x,y=y, step_size=0.001)
losses.append(loss)
if iteration % 1 == 0:
print("Iteration: {}".format(iteration))
print(loss)
输出
Iteration: 0
612601.7859402705
Iteration: 1
67456013215.98818
Iteration: 2
7427849474110884.0
Iteration: 3
8.179099502901393e+20
Iteration: 4
9.006330707513148e+25
Iteration: 5
9.917228672922966e+30
Iteration: 6
1.0920254505132042e+36
Iteration: 7
1.2024725981084638e+41
Iteration: 8
1.324090295064888e+46
Iteration: 9
1.4580083421516024e+51
Iteration: 10
1.60547085025467e+56
Iteration: 11
1.7678478362285333e+61
Iteration: 12
1.946647415292399e+66
Iteration: 13
2.1435307416407376e+71
Iteration: 14
2.3603265498975516e+76
Iteration: 15
2.599049318486855e+81
Iteration: 16
nan
Iteration: 17
nan
Iteration: 18
nan
Iteration: 19
nan
Iteration: 20
nan
Iteration: 21
nan
Iteration: 22
nan
Iteration: 23
nan
Iteration: 24
nan
Iteration: 25
nan
Iteration: 26
nan
Iteration: 27
nan
Iteration: 28
nan
Iteration: 29
nan
Iteration: 30
nan
Iteration: 31
nan
Iteration: 32
nan
Iteration: 33
nan
Iteration: 34
nan
Iteration: 35
nan
Iteration: 36
nan
Iteration: 37
nan
Iteration: 38
nan
Iteration: 39
nan
Iteration: 40
nan
Iteration: 41
nan
Iteration: 42
nan
Iteration: 43
nan
Iteration: 44
nan
Iteration: 45
nan
Iteration: 46
nan
Iteration: 47
nan
Iteration: 48
nan
Iteration: 49
nan
Iteration: 50
nan
Iteration: 51
nan
Iteration: 52
nan
Iteration: 53
nan
Iteration: 54
nan
Iteration: 55
nan
Iteration: 56
nan
Iteration: 57
nan
Iteration: 58
nan
Iteration: 59
nan
Iteration: 60
nan
Iteration: 61
nan
Iteration: 62
nan
Iteration: 63
nan
Iteration: 64
nan
Iteration: 65
nan
Iteration: 66
nan
Iteration: 67
nan
Iteration: 68
nan
Iteration: 69
nan
Iteration: 70
nan
Iteration: 71
nan
Iteration: 72
nan
Iteration: 73
nan
Iteration: 74
nan
Iteration: 75
nan
Iteration: 76
nan
Iteration: 77
nan
Iteration: 78
nan
Iteration: 79
nan
Iteration: 80
nan
Iteration: 81
nan
Iteration: 82
nan
Iteration: 83
nan
Iteration: 84
nan
Iteration: 85
nan
Iteration: 86
nan
Iteration: 87
nan
Iteration: 88
nan
Iteration: 89
nan
Iteration: 90
nan
Iteration: 91
nan
Iteration: 92
nan
Iteration: 93
nan
Iteration: 94
nan
Iteration: 95
nan
Iteration: 96
nan
Iteration: 97
nan
Iteration: 98
nan
Iteration: 99
nan
【问题讨论】：
标签： python numpy machine-learning linear-regression