1
我已经为python做了一个简单的梯度下降实现,它对大多数参数都能正常工作,但对于学习速率和迭代次数的某些参数,它会给我一个运行时错误。渐变下降运行时错误
RuntimeWarning:溢出double_scalars遇到
RuntimeWarning:在double_scalars
遇到无效值现在我假设,因为它得到一个地步,因为有B和M值太大而不能存储在内存中溢出错误,这个假设是否正确?
我该如何防止程序崩溃,因为主程序中的异常处理似乎不起作用,您能想出一种无异常处理的方式来防止错误在逻辑上?
def compute_error(points,b,m):
error = 0
for i in range(len(points)):
y = ponts[i][1]
x = points[i][0]
error += (y - (m*x + b))**2
return error/len(points)
def gradient_runner(points,LR,num_iter,startB=0,startM=0):
b = startB
m = startM
for i in range(num_iter):
b,m = step_gradient(points,b,m,LR)
return [b,m]
def step_gradient(points,b,m,LR):
b_gradient = 0
m_gradient = 0
N = float(len(points))
for i in range(len(points)):
x = points[i][0]
y = points[i][1]
b_gradient+= (-2/N)*(y - ((m*x)+b))
m_gradient+= (-2/N)*x*(y - ((m*x)+b))
## print "Value for b_gradient",b_gradient
## print "Value for b is ",b
## print "Value for learning rate is ",LR
new_b = b - (LR * b_gradient)
new_m = m - (LR * m_gradient)
return [new_b,new_m]
import numpy as np
a = np.array([[1,1],[4,2],[6,3],[8,4],[11,5],[12,6],[13,7],[16,8]])
b,m=gradient_runner(a,0.0001,1000) # These parameters work
# b,m=gradient_runner(a,0.1,10000) #Program Crashes
yguesses = [m * i + b for i in a[:,0]]
import matplotlib.pyplot as plt
guezz= yguesses
plt.scatter(a[:,0], a[:,1] ,color="green")
plt.plot(a[:,0],guezz,color="red")
plt.show()