在神经网络中使用一种热编码矢量输入标签 [英] Feeding Labels with one hot encoded vectors in Neural Network

问题描述

我正在尝试创建分类分类神经网络(NN) 我得到了具有169307行的数据集. 我的输出标签是[0,1,2] 我对它们进行了热编码，但无法使用Neural Nets对它们进行操作. 我遇到价值错误. 我想我在重塑目标"列时犯了错误. 我已经转换成清单l 这是我完整的解决方案代码.

I'm trying to create a Categorical classification Neural Network(NN) I have been given dataset which has 169307 rows. My output labels are [0,1,2] I one hot encoded them but I'm not able to operate on them using Neural Nets. I encounter Value Error. I guess I have made mistake in Reshaping my "target" column. I have converted into a list l Here is my complete Code to the Solution.

# coding: utf-8

# In[349]:

import pandas as pd
import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np
from sklearn import datasets
from sklearn import metrics
from sklearn import model_selection
from sklearn import preprocessing
from sklearn.preprocessing import OneHotEncoder


# In[382]:

df =pd.read_csv("train_data.csv")
num_labels = 3


# In[392]:

import numpy as np
nb_classes = 3
targets = np.array([0,1,2]).reshape(-1)
one_hot_targets = np.eye(nb_classes)[targets]
one_hot_targets


# In[420]:

target = df["target"]
feat=df.drop(['target','connection_id'],axis=1)
target[10]
l=[]
l=target.values.tolist()
l=np.array(l)
l[9]


# In[410]:




# In[394]:

logs_path="Server_attack"


# In[395]:

#Hyperparameters
batch_size=100
learning_rate=0.5
training_epochs=10


# In[396]:

X=tf.placeholder(tf.float32,[None,41])
Y_=tf.placeholder(tf.float32,[None,3])
lr=tf.placeholder(tf.float32)
XX=tf.reshape(X,[41,-1])  



# In[397]:

#5Layer Neural Network
L=200
M=100
N=60
O=30


# In[398]:

#Weights and Biases

W1=tf.Variable(tf.truncated_normal([41,L],stddev=0.1))
B1=tf.Variable(tf.ones([L]))
W2=tf.Variable(tf.truncated_normal([L,M],stddev=0.1))
B2=tf.Variable(tf.ones([M]))
W3=tf.Variable(tf.truncated_normal([M,N],stddev=0.1))
B3=tf.Variable(tf.ones([N]))
W4=tf.Variable(tf.truncated_normal([N,O],stddev=0.1))
B4=tf.Variable(tf.ones([O]))
W5=tf.Variable(tf.truncated_normal([O,3],stddev=0.1))
B5=tf.Variable(tf.ones([3]))               
               


# In[399]:

Y1=tf.nn.relu(tf.matmul(XX,W1)+B1)
Y2=tf.nn.relu(tf.matmul(Y1,W2)+B2)
Y3=tf.nn.relu(tf.matmul(Y2,W3)+B3)
Y4=tf.nn.relu(tf.matmul(Y3,W4)+B4)
Ylogits=tf.nn.relu(tf.matmul(Y4,W5)+B5)
Y=tf.nn.softmax(Ylogits)


# In[400]:

cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits,labels=Y_)
cross_entropy = tf.reduce_mean(cross_entropy)


# In[401]:

correct_prediction=tf.equal(tf.argmax(Y,1),tf.argmax(Y_,1))
accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))


# In[402]:

train_step=tf.train.AdamOptimizer(learning_rate).minimize(cross_entropy)


# In[403]:

#TensorBoard Parameters
tf.summary.scalar("cost",cross_entropy)
tf.summary.scalar("accuracy",accuracy)
summary_op=tf.summary.merge_all()


# In[404]:

init = tf.global_variables_initializer()
sess=tf.Session()
sess.run(init)


# In[417]:

with tf.Session() as sess:
    sess.run(init)
    writer = tf.summary.FileWriter(logs_path,graph=tf.get_default_graph())
    for epoch in range(training_epochs):
        batch_count=int(len(feature)/batch_size)
        for i in range(batch_count):

            
            #batch_x,batch_y=feature.iloc[i, :].values.tolist(),target[i]
            batch_x = np.expand_dims(np.array(feature.iloc[i, :].values.tolist()), axis=0)
            batch_y = np.expand_dims(l, axis=0)
            

           # batch_y = np.reshape(batch_y,(1, 3))



            _,summary = sess.run([train_step,summary_op],
                                 {X:batch_x,Y:batch_y,learning_rate:0.001}
                                )
            
            writer.add_summary(summary, epoch * batch_count + i)
        print("Epoch: ", epoch)

错误:

ValueError:无法为Tensor输入形状(1，169307)的值 'Softmax_16:0'，形状为'(41，3)'

ValueError: Cannot feed value of shape (1, 169307) for Tensor 'Softmax_16:0', which has shape '(41, 3)'

请指导我

推荐答案

您实际上没有进行转换.您只创建了一个3x3的单位矩阵one_hot_targets，但从未使用过它.结果，batch_y是df["target"]的数组:

You actually didn't do the conversion. You've only created a 3x3 identity matrix one_hot_targets, but never used it. As a result, batch_y is an array of df["target"]:

target = df["target"]
l = target.values.tolist()
l = np.array(l)
...
batch_y = np.expand_dims(l, axis=0)  # Has shape `(1, 169307)`!

您的batch_x似乎也不正确，但是feature并未在代码段中定义，因此我无法确切地说出它是什么.

Your batch_x also doesn't seem correct, but the feature is not defined in the snippet, so I can't say what exactly that is.

[更新] 如何进行一键编码:

from sklearn.preprocessing import OneHotEncoder

# Categorical target: 0, 1 or 2. The value is just an example
target = np.array([1, 2, 2, 1, 0, 2, 1, 1, 0, 2, 1])

target = target.reshape([-1, 1])      # add one extra dimension
encoder = OneHotEncoder(sparse=False)
encoder.fit(target)
encoded = encoder.transform(target)   # now it's one-hot: [N, 3]

这篇关于在神经网络中使用一种热编码矢量输入标签的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！