Python 神经网络输出所有1s
我正在尝试创建一个用于预测的ANN网络（即不只是分类为0或1） 我的代码如下：Python 神经网络输出所有1s,python,machine-learning,deep-learning,neural-network,prediction,Python,Machine Learning,Deep Learning,Neural Network,Prediction,我正在尝试创建一个用于预测的ANN网络（即不只是分类为0或1） 我的代码如下： import os import pandas as pd import numpy as np from sklearn import preprocessing from sklearn.preprocessing import MinMaxScaler from sklearn.model_selection import train_test_split data = pd.read_csv('ANNDat
import os
import pandas as pd
import numpy as np
from sklearn import preprocessing
from sklearn.preprocessing import MinMaxScaler
from sklearn.model_selection import train_test_split
data = pd.read_csv('ANNData.csv', skiprows=0, usecols=(1,2,3,4,5,6,7,8,9,11)) #importing data
data.interpolate(axis=0, method='linear', inplace=True) #interpolates = fills in missing values
data=data.values #converts to array
#scaling the data
min_max_scaler = preprocessing.MinMaxScaler()
scaled_data = min_max_scaler.fit_transform(data)
inputs=scaled_data[:,0:8]
output=scaled_data[:,8]
output = output.reshape(1095,1)
def sigmoid(x):
return 1/(1+(np.exp(-x)))
def sigmoid_der(x):
return (sigmoid(x)*(1-sigmoid(x)))
X_train, X_test, y_train, y_test = train_test_split(inputs, output, test_size=0.2)
synaptic_weights=2*np.random.random((8,1))-1
for iteration in range(10000):
input_layer= X_train
outputs = sigmoid(np.dot(input_layer, synaptic_weights))
error = outputs-y_train
adjustments = error*sigmoid_der(outputs)
synaptic_weights =np.dot(input_layer.T, adjustments)
print(outputs)
[0.69083156]
[0.07889126]
[0.3880597 ]
[0.3880597 ]
[0.23240938]
有什么建议吗？或者是添加更多层的唯一解决方案？假设您使用的是梯度下降，我认为您缺少学习速度 尝试设置
lr=1-e3#尝试从1-e2到1-e5的不同范围
调整=lr*误差*西格莫德（输出）
lr=1-e3#尝试从1-e2到1-e5的不同范围
调整=lr*误差*西格莫德（输出）
[1.]
[1.]
[1.]
[1.]