DAY65. Tensorflow Keras model (1)dnn model

keras binary

- Keras model : 이항분류기 
X변수 : 정규화(0~1)
Y변수 : one hot encoding(2진수)

from sklearn.datasets import load_iris #dataset 
from sklearn.model_selection import train_test_split #split 
from sklearn.preprocessing import minmax_scale #X변수 : 정규화(0~1)

from tensorflow.keras.utils import to_categorical #Y변수 : encoding 
from tensorflow.keras import Sequential #keras model 생성 
from tensorflow.keras.layers import Dense #DNN layer 구축

1. dataset load & 전처리 

X, y = load_iris(return_X_y=True)
X.shape #(150, 4) : 입력수=4

X변수 : 정규화 

X = minmax_scale(X[:100])
print(X)

Y변수 : one hot encoding

y_one = to_categorical(y[:100])
y_one.shape #(100, 2) : 출력수 : 2
print(y_one)

 

2. train_test_split

X_train, X_val, y_train, y_val = train_test_split(
    X, y_one, test_size=0.3, random_state=123)

 

3. keras model 

model = Sequential() #생성자 -> model 객체

4. DNN model layer 구축 
1층 - hidden1 : 뉴런 8개 
2층 - hidden2 : 뉴런 4개 
3층 - output  : 뉴런 2개

model.add(Dense(units=8, input_shape=(4,), activation='relu'))#1층 
model.add(Dense(units=4, activation='relu')) #2층 
model.add(Dense(units=2, activation='sigmoid')) #3층

model layer 확인 

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param=(in*out)+b   
=================================================================
dense (Dense)                (None, 8)                 40=(4*8)+8        
_________________________________________________________________
dense_1 (Dense)              (None, 4)                 36=(8*4)+4        
_________________________________________________________________
dense_2 (Dense)              (None, 2)                 10=(4*2)+2       
=================================================================
Total params: 86

 

5. model compile : 학습과정 설정(이항분류기)

model.compile(optimizer='adam', 
              loss='binary_crossentropy',
              metrics=['accuracy'])

optimizer : 최적화 알고리즘('adam', 'sgd')
loss : 손실함수('binary_crossentropy', 'categorical_crossentropy', 'mse')
metrics : 평가방법('accuracy', 'mae')

 

6. model training : train(70) vs val(30)

model.fit(x=X_train, y=y_train, #훈련셋 
          epochs=30, #반복학습 횟수 
          verbose=1, #출력여부 
          validation_data= (X_val, y_val)) #검증셋

Epoch 30/30
3/3 [==============================] - 0s 4ms/step - loss: 0.4898 - accuracy: 1.0000 - val_loss: 0.4971 - val_accuracy: 0.9667



7. model evaluation : val dataset

print('='*30)
print('model evaluation')
model.evaluate(x=X_val, y=y_val)

==============================
model evaluation 1/1 0s 2ms/step - loss: 0.4971 - accuracy: 0.9667
속도, 손실값, 정확률

 

 

 

 

 

keras dnn category

- Keras model : 다항분류기 
X변수 : 정규화(0~1)
Y변수 : one hot encoding(2진수)

from sklearn.datasets import load_iris #dataset 
from sklearn.model_selection import train_test_split #split 
from sklearn.preprocessing import minmax_scale #X변수 : 정규화(0~1)

from tensorflow.keras.utils import to_categorical #Y변수 : encoding 
from tensorflow.keras import Sequential #keras model 생성 
from tensorflow.keras.layers import Dense #DNN layer 구축

keras 내부 w,b변수 seed 적용 

import tensorflow as tf
import numpy as np 
import random as rd

tf.random.set_seed(123)
np.random.seed(123)
rd.seed(123)

1. dataset load & 전처리 

X, y = load_iris(return_X_y=True)

X.shape #(150, 4) : 입력수=4
print(y) #0~2 : class

X변수 : 정규화 

X = minmax_scale(X) #[수정]
print(X)

Y변수 : class -> one hot encoding

y_one = to_categorical(y) #[수정]
y_one.shape #(150, 3) : 출력수 : 3
print(y_one)

2. train_test_split

X_train, X_val, y_train, y_val = train_test_split(
    X, y, test_size=0.3, random_state=123) #y_one -> y

 

3. keras model 

model = Sequential() #생성자 -> model 객체

4. DNN model layer 구축 : [수정]
1층 - hidden1 : 뉴런 8개 -> 12개
2층 - hidden2 : 뉴런 4개 -> 6개 
3층 - output  : 뉴런 2개 -> 3개 

model.add(Dense(units=12, input_shape=(4,), activation='relu'))#1층 
model.add(Dense(units=6, activation='relu')) #2층 
model.add(Dense(units=3, activation='softmax')) #3층 : [수정]

model layer 확인 

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_42 (Dense)             (None, 12)                60        
_________________________________________________________________
dense_43 (Dense)             (None, 6)                 78        
_________________________________________________________________
dense_44 (Dense)             (None, 3)                 21        
============================================================
Total params: 159

 

5. model compile : 학습과정 설정(다항분류기) 

model.compile(optimizer='adam', 
              #loss='categorical_crossentropy', #[수정] y : one_hot encoding 
              loss='sparse_categorical_crossentropy', #y : class(0,1,2)
              metrics=['accuracy'])

optimizer : 최적화 알고리즘('adam', 'sgd')
loss : 손실함수('binary_crossentropy', 'categorical_crossentropy', 'mse')
metrics : 평가방법('accuracy', 'mae')

손실함수 : y변수 type에 의해서 결정 
y : one_hot encoding(100,010,001) -> lost = 'categorical_crossentropy'
y : class(0, 1, 2) -> lost = 'sparse_categorical_crossentropy'

 

6. model training : train(70) vs val(30)

model.fit(x=X_train, y=y_train, # 훈련셋 
          epochs=200, #반복학습 횟수 [수정]
          verbose=1, #출력여부 
          validation_data= (X_val, y_val)) #검증셋

 

 

#7. model evaluation : val dataset 

print('='*30)
print('model evaluation')
model.evaluate(x=X_val, y=y_val)

epochs=100 : 0s 499us/step - loss: 0.3624 - accuracy: 0.8667
epochs=200 : 0s 499us/step - loss: 0.1601 - accuracy: 0.9556

 

8. model save & load : hdf5 파일 형식 

from tensorflow.keras.models import load_model #model load 
from sklearn.metrics import accuracy_score #model 평가

1) model save

model.save(filepath='keras_model_iris.h5')

2) model load 

new_model = load_model(filepath='keras_model_iris.h5')

9. model test : test set 

X_train, X_test, y_train, y_test = train_test_split(
    X, y_one, test_size=0.5, random_state=123)

y_pred = new_model.predict(X_test) #new dataset 

y_pred #확률예측 : softmax 함수 
y_pred.shape #(75, 3)

y_pred : 확률예측 -> class 변경 

y_pred = tf.argmax(y_pred, axis=1) #행단위 
#y_true : one-hot encoding -> class 변경 
y_true = tf.argmax(y_test, axis=1) #행단위

acc = accuracy_score(y_true, y_pred)
print('accuracy =',acc) #accuracy = 0.9733333333333334

 

 

 

 

 

keras dnn category learning rate

* keras dnn category 참고 

- Keras model : 다항분류기 
X변수 : 정규화(0~1)
Y변수 : one hot encoding(2진수)
optimizer='adam' -> optimizer=Adam(learning_rate=0.01)

from sklearn.datasets import load_iris #dataset 
from sklearn.model_selection import train_test_split #split 
from sklearn.preprocessing import minmax_scale #X변수 : 정규화(0~1)

from tensorflow.keras.utils import to_categorical #Y변수 : encoding 
from tensorflow.keras import Sequential #keras model 생성 
from tensorflow.keras.layers import Dense #DNN layer 구축

keras 내부 w,b변수 seed 적용 

import tensorflow as tf
import numpy as np 
import random as rd

tf.random.set_seed(123)
np.random.seed(123)
rd.seed(123)

 

 

1. dataset load & 전처리 

X, y = load_iris(return_X_y=True)

X.shape #(150, 4) : 입력수=4

X변수 : 정규화 

X = minmax_scale(X) 
print(X)

Y변수 : one hot encoding

y_one = to_categorical(y) 
y_one.shape #(150, 3) : 출력수 : 3
print(y_one)

 

2. train_test_split

X_train, X_val, y_train, y_val = train_test_split(
    X, y_one, test_size=0.3, random_state=123)

 

3. keras model 

model = Sequential() #생성자 -> model 객체

4. DNN model layer 구축 
1층 - hidden1 : 뉴런 8개 -> 12개
2층 - hidden2 : 뉴런 4개 -> 6개 
3층 - output  : 뉴런 2개 -> 3개 

model.add(Dense(units=12, input_shape=(4,), activation='relu'))#1층 
model.add(Dense(units=6, activation='relu')) #2층 
model.add(Dense(units=3, activation='softmax')) #3층

model layer 확인 

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_42 (Dense)             (None, 12)                60        
_________________________________________________________________
dense_43 (Dense)             (None, 6)                 78        
_________________________________________________________________
dense_44 (Dense)             (None, 3)                 21        
============================================================
Total params: 159

 

5. model compile : 학습과정 설정(다항분류기) - [수정]

from tensorflow.keras import optimizers #Adam 사용 

#default : learning_rate=0.001
model.compile(optimizer=optimizers.Adam(learning_rate=0.01), #[수정]
              loss='categorical_crossentropy', 
              metrics=['accuracy'])

optimizer : 최적화 알고리즘('adam', 'sgd')
loss : 손실함수('binary_crossentropy', 'categorical_crossentropy', 'mse')
metrics : 평가방법('accuracy', 'mae')

 

6. model training : train(70) vs val(30)

model.fit(x=X_train, y=y_train, # 훈련셋 
          epochs=200, #반복학습 횟수 
          verbose=1, #출력여부 
          validation_data= (X_val, y_val)) #검증셋

7. model evaluation : val dataset 

print('='*30)
print('model evaluation')
model.evaluate(x=X_val, y=y_val)

epochs=100 : 0s 499us/step - loss: 0.3624 - accuracy: 0.8667
epochs=200(lr=0.001) : 0s 499us/step - loss: 0.1601 - accuracy: 0.9556
epochs=200(lr=0.01) : 0s 968us/step - loss: 0.0724 - accuracy: 0.9778

 

 

8. model save & load : hdf5 파일 형식 

from tensorflow.keras.models import load_model #model load 
from sklearn.metrics import accuracy_score #model 평가

1) model save

model.save(filepath='keras_model_iris.h5')

2) model load 

new_model = load_model(filepath='keras_model_iris.h5')

9. model test : test set 

X_train, X_test, y_train, y_test = train_test_split(
    X, y_one, test_size=0.5, random_state=123)

y_pred = new_model.predict(X_test) #new dataset 

y_pred #확률예측 : softmax 함수 
y_pred.shape #(75, 3)

y_pred : 확률예측 -> class 변경 

y_pred = tf.argmax(y_pred, axis=1) #행단위 
#y_true : one-hot encoding -> class 변경 
y_true = tf.argmax(y_test, axis=1) #행단위

acc = accuracy_score(y_true, y_pred)
print('accuracy =',acc)

accuracy = 0.9733333333333334 -> learning_rate=0.001
accuracy = 0.9866666666666667 -> learning_rate=0.01

 

 

 

 

 

keras mnist batch

1. Mnist dataset 다항분류기 
   - X(images) : 2d(28x28) = 1d(784 pixel)
   - Y(label) : 0~9(10진수)

2. Full batch vs Mini batch 
   - Full batch : 전체 훈련셋 공급 
   - Mini batch : 훈련셋 분할 공급(size 큰 경우)

from tensorflow.keras.datasets import mnist # mnist load 
from tensorflow.keras.utils import to_categorical # Y변수 : encoding 
from tensorflow.keras import Sequential # keras model 생성 
from tensorflow.keras.layers import Dense # DNN layer 구축

import tensorflow as tf
import numpy as np 
import random as rd

tf.random.set_seed(123)
np.random.seed(123)
rd.seed(123)
import time #학습 소요 시간 측정

1. mnist dataset load 

(x_train, y_train), (x_val, y_val) = mnist.load_data() #(images, labels)

images : X변수 

x_train.shape #(60000, 28, 28) - (size, h, w) : 2d 제공 
x_val.shape #(10000, 28, 28)

x_train[0] #0~255
x_train.max() #255

labels : y변수 

y_train.shape #(60000,)
y_train[0] #5

2. X,y변수 전처리 
1) X변수 : 정규화 & reshape(2d -> 1d)

x_train = x_train / 255. #정규화 
x_val = x_val / 255.

x_train[0]

reshape(2d -> 1d)

x_train = x_train.reshape(-1, 784) #(60000, 28*28)
x_val = x_val.reshape(-1, 784) #(10000, 28*28)

2) y변수 : class(10진수) -> one-hot encoding(2진수)

y_train = to_categorical(y_train)
y_val = to_categorical(y_val)

전처리 확인 

x_train.shape # (60000, 784)
y_train[0] #[0., 0., 0., 0., 0., 1., 0., 0., 0., 0.] - 5
y_train.shape #(60000, 10)

start_time = time.time() #소요 시간 체크

3. keras model

model = Sequential()

4. DNN model layer 구축 
hidden layer1 : w[784, 128]

model.add(Dense(units=128, input_shape=(784,), activation='relu')) #1층

hidden layer2 : w[128, 64]

model.add(Dense(units=64, activation='relu')) #2층

hidden layer3 : w[64, 32]

model.add(Dense(units=32, activation='relu')) #3층

output layer : w[32, 10]

model.add(Dense(units=10, activation='softmax')) #4층

model layer 확인 

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_8 (Dense)              (None, 128)               100480=784x128+128    
_________________________________________________________________
dense_9 (Dense)              (None, 64)                8256      
_________________________________________________________________
dense_10 (Dense)             (None, 32)                2080      
_________________________________________________________________
dense_11 (Dense)             (None, 10)                330       
=================================================================
Total params: 111,146

 

5. model compile : 학습과정 설정(다항분류기) 

model.compile(optimizer='adam', #default : learning_rate=0.001
              loss='categorical_crossentropy', 
              metrics=['accuracy'])

6. model training : train(70) vs val(30)

model.fit(x=x_train, y=y_train, #훈련셋 
          epochs=10, #반복학습 횟수 : 60000 * 10 = 600,000 -> full batch
          batch_size=100, #1epoch(100 * 600) * 10 = 600,000 -> mini batch 
          verbose=1, #출력여부 
          validation_data= (x_val, y_val)) #검증셋

stop_time = time.time() - start_time 

print('소요시간 : ', stop_time)

full batch 
accuracy: 0.9923 - val_loss: 0.0873 - val_accuracy: 0.9793
소요시간 :  25.037985801696777

mini batch 
accuracy: 0.9933 - val_loss: 0.0916 - val_accuracy: 0.9739
소요시간 :  11.333117961883545



7. model evaluation : val dataset 

print('model evaluation')
model.evaluate(x=x_val, y=y_val)
#loss: 0.0916 - accuracy: 0.9739

 

 

 

 

 

keras_mnist_flatten

1. Mnist dataset 다항분류기 
   - X(images) : 2d(28x28) -> model 공급 
   - Y(label) : 0~9(10진수)

2. Full batch vs Mini batch 
   - Full batch : 전체 훈련셋 공급 
   - Mini batch : 훈련셋 분할 공급(size 큰 경우)

3. Flatten layer : 차원일치(input : 2d -> 1d) 

from tensorflow.keras.datasets import mnist #mnist load 
from tensorflow.keras.utils import to_categorical #Y변수 : encoding 
from tensorflow.keras import Sequential #keras model 생성 
from tensorflow.keras.layers import Dense, Flatten  #DNN layer 구축

 

keras 내부 w,b변수 seed 적용 

import tensorflow as tf
import numpy as np 
import random as rd

tf.random.set_seed(123)
np.random.seed(123)
rd.seed(123)
import time #학습 소요 시간 측정

 

1. mnist dataset load 

(x_train, y_train), (x_val, y_val) = mnist.load_data() #(images, labels)

images : X변수 

x_train.shape #(60000, 28, 28) - (size, h, w) : 2d 제공 
x_val.shape #(10000, 28, 28)

x_train[0] #0~255
x_train.max() #255

labels : y변수 

y_train.shape #(60000,)
y_train[0] #5

2. X,y변수 전처리 
1) X변수 : 정규화 

x_train = x_train / 255. #정규화 
x_val = x_val / 255.

x_train[0]

 

[생략] reshape(2d -> 1d)
x_train = x_train.reshape(-1, 784) # (60000, 28*28)
x_val = x_val.reshape(-1, 784) # (10000, 28*28)


2) y변수 : class(10진수) -> one-hot encoding(2진수)

y_train = to_categorical(y_train)
y_val = to_categorical(y_val)

 

전처리 확인 

x_train.shape # (60000, 784)
y_train[0] # [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.] - 5
y_train.shape # (60000, 10)

start_time = time.time() #소요 시간 체트

3. keras model

model = Sequential()

4. DNN model layer 구축 

input_shape = (28, 28) #[추가] 2d image 공급

[추가] flatten layer  

model.add(Flatten(input_shape=input_shape))  #2d -> 1d(784)

hidden layer1 : w[784, 128]

model.add(Dense(units=128, input_shape=(784,), activation='relu')) #1층

 

hidden layer2 : w[128, 64]

model.add(Dense(units=64, activation='relu')) #2층

hidden layer3 : w[64, 32]

model.add(Dense(units=32, activation='relu')) #3층

output layer : w[32, 10]

model.add(Dense(units=10, activation='softmax')) #4층

model layer 확인 

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_8 (Dense)              (None, 128)               100480=784x128+128    
_________________________________________________________________
dense_9 (Dense)              (None, 64)                8256      
_________________________________________________________________
dense_10 (Dense)             (None, 32)                2080      
_________________________________________________________________
dense_11 (Dense)             (None, 10)                330       
=================================================================
Total params: 111,146



5. model compile : 학습과정 설정(다항분류기) 

model.compile(optimizer='adam', #default : learning_rate=0.001
              loss='categorical_crossentropy', 
              metrics=['accuracy'])

6. model training : train(70) vs val(30)

model.fit(x=x_train, y=y_train, #훈련셋 
          epochs=10, #반복학습 횟수 : 60000 * 10 = 600,000 -> full batch
          batch_size=100, #1epoch(100 * 600) * 10 = 600,000 -> mini batch 
          verbose=1, #출력여부 
          validation_data= (x_val, y_val)) #검증셋

stop_time = time.time() - start_time 

print('소요시간 : ', stop_time)

full batch 
accuracy: 0.9923 - val_loss: 0.0873 - val_accuracy: 0.9793
소요시간 :  25.037985801696777

mini batch 
accuracy: 0.9933 - val_loss: 0.0916 - val_accuracy: 0.9739
소요시간 :  11.333117961883545



7. model evaluation : val dataset 

print('model evaluation')
model.evaluate(x=x_val, y=y_val) #loss: 0.0916 - accuracy: 0.9739