李宏毅老师的demo
数据准备
# 假设x_data和y_data都有10笔,分别代表宝可梦进化前后的cp值
x_data=[338.,333.,328.,207.,226.,25.,179.,60.,208.,606.]
y_data=[640.,633.,619.,393.,428.,27.,193.,66.,226.,1591.]
# 这里采用最简单的linear model:y_data=b+w*x_data
# 我们要用gradient descent把b和w找出来
计算梯度微分的函数getGrad()
# 计算梯度微分的函数getGrad()
def getGrad(b,w):
# initial b_grad and w_grad
b_grad=0.0
w_grad=0.0
for i in range(10):
b_grad+=(-2.0)*(y_data[i]-(b+w*x_data[i]))
w_grad+=(-2.0*x_data[i])*(y_data[i]-(b+w*x_data[i]))
return (b_grad,w_grad)
引入需要的库
import matplotlib
import matplotlib.pyplot as plt
matplotlib.use('Agg')
%matplotlib inline
import random as random
import numpy as np
import csv
准备好b,w,loss的图像数据
# 生成一组b和w的数据图,方便给gradient descent的过程做标记
x = np.arange(-200,-100,1) # bias
y = np.arange(-5,5,0.1) # weight
Z = np.zeros((len(x),len(y))) # color
X,Y = np.meshgrid(x,y)
for i in range(len(x)):
for j in range(len(y)):
b = x[i]
w = y[j]
# Z[j][i]存储的是loss
Z[j][i] = 0
for n in range(len(x_data)):
Z[j][i] = Z[j][i] + (y_data[n] - (b + w * x_data[n]))**2
Z[j][i] = Z[j][i]/len(x_data)
使用Adagrad
# 这里给b和w不同的learning rate
# y_data = b + w * x_data
b = -120 # initial b
w = -4 # initial w
lr = 1 # learning rate 放大10倍
iteration = 100000 # 这里直接规定了迭代次数,而不是一直运行到b_grad和w_grad都为0(事实证明这样做不太可行)
# store initial values for plotting,我们想要最终把数据描绘在图上,因此存储过程数据
b_history = [b]
w_history = [w]
lr_b = 0
lr_w = 0
# iterations
for i in range(iteration):
# get new b_grad and w_grad
b_grad,w_grad=getGrad(b,w)
# get the different learning rate for b and w
lr_b = lr_b + b_grad ** 2
lr_w = lr_w + w_grad ** 2
# 这一招叫做adagrad,之后会详加解释
# update b and w with new learning rate
b -= lr / np.sqrt(lr_b) * b_grad
w -= lr / np.sqrt(lr_w) * w_grad
#store parameters for plotting
b_history.append(b)
w_history.append(w)
# output the b w b_grad w_grad
# print("b: "+str(b)+"\t\t\t w: "+str(w)+"\n"+"b_grad: "+str(b_grad)+"\t\t w_grad: "+str(w_grad)+"\n")
# output the final function and its error
print("the function will be y_data="+str(b)+"+"+str(w)+"*x_data")
error=0.0
for i in range(10):
print("error "+str(i)+" is: "+str(np.abs(y_data[i]-(b+w*x_data[i])))+" ")
error+=np.abs(y_data[i]-(b+w*x_data[i]))
average_error=error/10
print("the average error is "+str(average_error))
# plot the figure
plt.contourf(x,y,Z,50,alpha=0.5,cmap=plt.get_cmap('jet'))
plt.plot([-188.4],[2.67],'x',ms=12,markeredgewidth=3,color='orange')
plt.plot(b_history,w_history,'o-',ms=3,lw=1.5,color='black')
plt.xlim(-200,-100)
plt.ylim(-5,5)
plt.xlabel(r'$b$',fontsize=16)
plt.ylabel(r'$w$',fontsize=16)
plt.show()