Python Tensorflow Softmax

Softmax : 다중분류

예제

# softmax : 다중분류

# 분류 클래스 수만큼의 값이 산출됨

# 산출된 값을 합하면 1이 됨

#!pip install tensorflow

import tensorflow as tf

from tensorflow.keras.models import Sequential

from tensorflow.keras.layers import Dense

np.random.seed(0)

tf.random.set_seed(3)

from sklearn.model_selection import train_test_split

import pandas as pd

df = pd.read_csv('housing.csv',delim_whitespace=True,header=None)

0	1	2	3	4	5	6	7	8	9	10	11	12	13
0	0.00632	18.0	2.31	0	0.538	6.575	65.2	4.0900	1	296.0	15.3	396.90	4.98	24.0
1	0.02731	0.0	7.07	0	0.469	6.421	78.9	4.9671	2	242.0	17.8	396.90	9.14	21.6
2	0.02729	0.0	7.07	0	0.469	7.185	61.1	4.9671	2	242.0	17.8	392.83	4.03	34.7
3	0.03237	0.0	2.18	0	0.458	6.998	45.8	6.0622	3	222.0	18.7	394.63	2.94	33.4
4	0.06905	0.0	2.18	0	0.458	7.147	54.2	6.0622	3	222.0	18.7	396.90	5.33	36.2
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
501	0.06263	0.0	11.93	0	0.573	6.593	69.1	2.4786	1	273.0	21.0	391.99	9.67	22.4
502	0.04527	0.0	11.93	0	0.573	6.120	76.7	2.2875	1	273.0	21.0	396.90	9.08	20.6
503	0.06076	0.0	11.93	0	0.573	6.976	91.0	2.1675	1	273.0	21.0	396.90	5.64	23.9
504	0.10959	0.0	11.93	0	0.573	6.794	89.3	2.3889	1	273.0	21.0	393.45	6.48	22.0
505	0.04741	0.0	11.93	0	0.573	6.030	80.8	2.5050	1	273.0	21.0	396.90	7.88	11.9

506 rows × 14 columns

dataset = df.values

X = dataset[:,:13]

y = dataset[:,13]

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.3,random_state=0)

X_train.shape # (354, 13)

X_test.shape # (152, 13)

y_train.shape # (354,)

y_test.shape # (152,)

model = Sequential()

model.add(Dense(30,input_dim=13, activation='relu'))

model.add(Dense(20,activation='relu')) # 히든레이어

model.add(Dense(10,activation='relu')) # 히든레이어

model.add(Dense(15,activation='relu')) # 히든레이어

model.add(Dense(1)) # 결과

model.summary()

Model: "sequential_1"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense_2 (Dense)             (None, 30)                420       
                                                                 
 dense_3 (Dense)             (None, 20)                620       
                                                                 
 dense_4 (Dense)             (None, 10)                210       
                                                                 
 dense_5 (Dense)             (None, 15)                165       
                                                                 
 dense_6 (Dense)             (None, 1)                 16        
                                                                 
=================================================================
Total params: 1,431
Trainable params: 1,431
Non-trainable params: 0
_________________________________________________________________

# Loss함수 'adam' 사용

model.compile(loss='mse', optimizer='adam')

# 학습내역

history = model.fit(X_train,y_train,epochs=200,batch_size=10) # epoch : 반복횟수, batch_size : 10개의 입력을 처리하고 손실함수를 기반으로 가중치 조정, 현재 1개 입력은 13의 input을 말한다

# 인공신경망이 계산한 집값

model.predict(X_test).flatten()

# 학습내역

history.history['loss']

# 학습내역 그래프

plt.plot(history.history['loss'])

plt.show()

예제2

# Tensorflow를 사용한 지도학습, 다중분류

# 회귀(Regression), 분류(Classification)

# 회귀 : 출력 레이어의 뉴런의 수가 1개

# 분류 : 출력 레이어의 뉴런의 수가 클래스의 수와 같이 설정

# 출력함수 : 회귀(linear), 분류(softmax)

# 손실함수 : crossentropy(Log Loss)

# binary_crossentropy, categorical_crossentropy, sparse_crossentropy

# 회원분류, 비지도학습

# 이미지 분류 : CNN, 지도학습

# 오브젝트 탐지

# Tensorflow를 이용한 다중분류의 예

# iris 품종 분류

# 분류 클래스의 수 : 3

# feature의 수 : 4

import tensorflow as tf

from tensorflow.keras.models import Sequential

from tensorflow.keras.layers import Dense

import pandas as pd

import numpy as np

from sklearn.datasets import load_iris

iris = load_iris()

iris.keys() # 'data', 'target', 'target_names'

data = iris['data']

data.shape # (150, 4)

target = iris['target']

target.shape # (150,)

target_names = iris['target_names']

target_names.shape # (3,)

from sklearn.model_selection import train_test_split

X_train,X_test,y_train,y_test = train_test_split(data, target, test_size=.2, random_state=0)

X_train.shape # (120, 4)

X_test.shape # (30, 4)

y_train.shape # (120,)

y_test.shape # (30,)

# 신경망

model = Sequential() # 빈 신경망

model.add(Dense(10,input_dim=4,activation='relu'))

model.add(Dense(20,activation='relu'))

model.add(Dense(30,activation='relu'))

model.add(Dense(20,activation='relu'))

model.add(Dense(10,activation='relu'))

model.add(Dense(3,activation='softmax')) # 다중분류는 Softmax

model.summary()

Model: "sequential_3"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense_6 (Dense)             (None, 10)                50        
                                                                 
 dense_7 (Dense)             (None, 20)                220       
                                                                 
 dense_8 (Dense)             (None, 30)                630       
                                                                 
 dense_9 (Dense)             (None, 20)                620       
                                                                 
 dense_10 (Dense)            (None, 10)                210       
                                                                 
 dense_11 (Dense)            (None, 3)                 33        
                                                                 
=================================================================
Total params: 1,763
Trainable params: 1,763
Non-trainable params: 0
_________________________________________________________________

model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy']) # optimizer : 경사하강법의 종류, loss : 손실함수, 'sparse' : OneHotEncoding 방식으로 처리, metrics : 측정도구

# 학습

hist = model.fit(X_train, y_train, epochs=500, batch_size=10, validation_split=0.05) # epochs : 500번 반복, batch_size : input 10번 당 가중치 조정, validation_split : 

hist.history.keys() # dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])

loss = hist.history['loss']

acc = hist.history['accuracy']

vloss = hist.history['val_loss']

vacc = hist.history['val_accuracy']

import matplotlib.pyplot as plt

plt.figure(figsize=(5,2))

plt.plot(loss,'r--',label='loss')

plt.plot(acc,'g.',label='accuracy')

plt.plot(vloss,'y--',label='val_loss')

plt.plot(vacc,'b.',label='val_accuracy')

plt.legend()

plt.show()

model.evaluate(X_test, y_test)

1/1 [==============================] - 0s 29ms/step - loss: 0.0353 - accuracy: 1.0000

# 추정

pred = model.predict(X_test)

pred.shape # (30, 3)

pred

sum(pred[0])

pred[0].argmax() # 2

target_names[2] # 'virginica'

예제3

# 신경망을 이용한 회귀 실습

# 손실함수 : mse

# Optimizer : adam

# 활성함수 : activation='relu'

# 출력함수 : activation='linear'

# 출력 레이어의 뉴런 수 : 1

def get_y(x):

    return 4 * x + 7

x = np.random.randint(-100, 100, 500)

y = get_y(x)

y.shape # (500,)

X = x.reshape(-1,1)

X_train,X_test,y_train,y_test = train_test_split(X, y, test_size=.2, random_state=0)

X_train.shape, X_test.shape, y_train.shape, y_test.shape # ((400,), (100,), (400,), (100,))

model = Sequential() # 빈 신경망

model.add(Dense(1,input_dim=1)) # 기본 activation='linear' 생략가능

model.add(Dense(10))

model.add(Dense(15))

model.add(Dense(1))

model.summary()

model.compile(loss='mse', optimizer='adam',metrics=['mae']) # mae(Mean Absolute Error) : 오차를 가지고 정확도를 계산

hist = model.fit(X_train,y_train,epochs=100,batch_size=10,validation_split=.1)

hist.history.keys() # dict_keys(['loss', 'mae', 'val_loss', 'val_mae'])

loss = hist.history['loss']

mae = hist.history['mae']

val_loss = hist.history['val_loss']

val_mae = hist.history['val_mae']

plt.plot(loss,label='loss')

plt.plot(mae,label='mae')

plt.plot(val_loss,label='val_loss')

plt.plot(val_mae,label='val_mae')

plt.legend()

plt.show()

# 테스트

get_y(-50) # -193

model.predict([[-50]]) # -193.00005

model.evaluate(X_test, y_test) # [7.292737524977611e-10, 2.1238327462924644e-05]