Keras入门-搭建BP网络识别手写字符

最新推荐文章于 2024-01-18 23:43:54 发布

Yogaht

最新推荐文章于 2024-01-18 23:43:54 发布

阅读量812

点赞数 1

CC 4.0 BY-SA版权

分类专栏：神经网络文章标签： keras入门 minst手写字符神经网络 python

本文链接：https://blog.youkuaiyun.com/Yogaht/article/details/85051466

神经网络专栏收录该内容

2 篇文章

订阅专栏

本文详细介绍了一个基于BP神经网络的数据分类项目。从数据读取、预处理到模型搭建、训练及评估，全程实战演示。利用Python的pandas和sklearn库进行数据处理，Keras框架实现神经网络模型，通过标准化、one-hot编码等技术提升模型性能。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

代码

#读取数据
import pandas as pd

def retrieveData():
    trainingdata=pd.read_csv("training-data.csv").as_matrix()
    testdata=pd.read_csv("test-data.csv").as_matrix()
    
    return trainingdata,testdata

trainingdata,testdata=retrieveData()
print(trainingdata.shape)

print(testdata.shape)

print(trainingdata[:3])

输出：

(7493, 17)
(3497, 17)
[[ 0 89 27 100 42 75 29 45 15 15 37 0 69 2 100 6 2]
[ 0 57 31 68 72 90 100 100 76 75 50 51 28 25 16 0 1]
[ 0 100 7 92 5 68 19 45 86 34 100 45 74 23 67 0 4]]

#数据分割
def separateFeaturesAndCategories(trainingdata,testdata):
    
    trainingFeatures=trainingdata[:,:-1]
    trainingCategories=trainingdata[:,-1:]
    testFeatures=testdata[:,:-1]
    testCategories=testdata[:,-1:]
    
    return trainingFeatures,trainingCategories,testFeatures,testCategories

trainingFeatures,trainingCategories,testFeatures,testCategories=separateFeaturesAndCategories(trainingdata,testdata)

print(trainingFeatures[:3])
print(trainingCategories[:3].T)

输出：

[[ 0 89 27 100 42 75 29 45 15 15 37 0 69 2 100 6]
[ 0 57 31 68 72 90 100 100 76 75 50 51 28 25 16 0]
[ 0 100 7 92 5 68 19 45 86 34 100 45 74 23 67 0]]
[[2 1 4]]

#数据归一化
from sklearn.preprocessing import StandardScaler

def scaleData(trainingFeatures,testFeatures):
    scaler=StandardScaler()
    scaler.fit(trainingFeatures)
    scaledtrainingFeatures=scaler.transform(trainingFeatures)
    scaledtestFeatures=scaler.transform(testFeatures)
    
    return scaledtrainingFeatures,scaledtestFeatures
scaledtrainingFeatures,scaledtestFeatures=scaleData(trainingFeatures,testFeatures)

print(scaledtrainingFeatures[:1])

输出：

[[-1.12187758 0.25657834 -0.49541212 0.87125674 -0.25417164 0.36353124
-0.73260178 0.01286466 -1.25134391 -0.6944943 -0.6576261 -1.30131455
0.63319745 -1.01733507 1.28059647 -0.63002209]]

#对于多分类问题,将标签进行one-hot编码,采用to_categorical函数.
from keras.utils.np_utils import to_categorical
trainingCategories=to_categorical(trainingCategories, num_classes=10)
testCategories=to_categorical(testCategories, num_classes=10)

输出：

[[8]
[8]
[9]]
[[0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
[0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
[0. 0. 0. 0. 0. 0. 0. 0. 0. 1.]]

#搭建BP网络
from keras.models import Sequential
from keras.layers import Activation,Dense
model=Sequential()
#model=Sequential([Dense(16,input_dim=16),Activation('relu'),Dense(10),Activation('softmax')])

model.add(Dense(16,input_shape=(16,)))
model.add(Activation('relu'))
#model.add(Dropout(0.5))

model.add(Dense(16,activation='relu'))


model.add(Dense(10))
model.add(Activation('softmax'))

model.compile(loss='categorical_crossentropy',optimizer='adadelta',metrics=['accuracy'])  
#softmax对应损失函数categorical_crossentropy,多分类问题
model.fit(scaledtrainingFeatures,trainingCategories,epochs=20,batch_size=20)


accuracy=model.evaluate(scaledtestFeatures,testCategories,verbose=0)  
print('\nTest accuracy:',accuracy[1])
print('\nTest score:',accuracy[0])

Epoch 19/20
7493/7493 [==============================] - 6s 855us/step - loss: 1.1042 - acc: 0.6349
Epoch 20/20
7493/7493 [==============================] - 6s 845us/step - loss: 1.0671 - acc: 0.6434

Test accuracy: 0.8581641407431553

Test score: 0.4339949927876805

附：如何调参？

https://www.cnblogs.com/mrxsc/articles/6266584.html