Question

＆＃34;我必须再次提出这个问题，因为以前没有适当的帮助。＆＃34;

我最近学会了批量梯度下降算法并尝试在Python中实现它。我给出了一个非随机的数据集。当我尝试运行下面的代码时，该过程在3次迭代后收敛，但是出现了很大的错误。有人能以正确的方式指导我吗？样本数据集:(原始数据集长度为600。）

6203.75 1   173.8   43.6    0.0 183.0
6329.75 1   115.0   60.1    0.0 236.2
5830.75 1   159.5   94.1    21.0    275.8
4061.75 1   82.5    45.0    11.0    75.7
3311    1   185.5   46.1    4.0 0.0
4349.75 1   169.5   40.3    5.0 73.5
5695.25 1   138.5   68.9    6.0 204.2
5633.5  1   50.0    117.3   4.0 263.9

第一列是输出。第二列是常数值。休息是特色。

谢谢

data = open('Data_trial.txt','r')
import time
lines=data.readlines()
dataSet=[]
for line in lines:                       
    dataSet.append(line.split())
original_output=[]
features=[]
for i in range(0,len(dataSet)):
    features.append([])
predict=[]
grad=[]
weights=[0,0,0,0,0]
learning_factor=0.01
for i in range(0,len(dataSet)):
    for j in range(0,len(dataSet[i])):
        if j==0:
            original_output.append(float(dataSet[i][j]))
        else:
            features[i].append(float(dataSet[i][j]))
def prediction(predict,weights,original_output,features):
    for count in range(0,len(original_output)):
        predict.append(sum(weights[i]*features[count][i] for i in range(0,len(features[count]))))
    print("predicted values",predict)

def gradient(predict,grad,original_output,features):
    for count in range(0,len(weights)):
        grad.append(sum((predict[i]-original_output[i])*features[i][count]
                        for i in range(0,len(original_output))))
    print("Gradient values",grad)
def weights_update(grad,learning_factor,weights):
    for i in range(0,len(weights)):
        weights[i]-=learning_factor*grad[i]
    print("Updated weights",weights)    
if __name__=="__main__":
    while True:
        prediction(predict,weights,original_output,features)
        gradient(predict,grad,original_output,features)
        weights_update(grad,learning_factor,weights)
        time.sleep(1)
        predict=[]
        grad=[]
        print()

python中的批量梯度下降算法实现

0 个答案: