[AI] 다변수 선형 회귀

Sequential API

 

csv.csv

0.00MB

import tensorflow as tf
import numpy as np

from numpy import genfromtxt
from tensorflow.keras.models import Sequential, Model
from tensorflow.keras.layers import Flatten, Dense, Input
from tensorflow.keras.optimizers import SGD, Adam

my_data = genfromtxt('csv.csv', delimiter=',')

print(my_data)

x_data = my_data[ : , :-1]
t_data = my_data[ : , [-1]]

print('loaded_data.shape = ', my_data.shape)
print('x_data.shape = ', x_data.shape, ', t_data.shape = ', t_data.shape)

loaded_data.shape =  (15, 4)
x_data.shape =  (15, 3) , t_data.shape =  (15, 1)

# Linear Regression 을 keras 이용하여 생성
model = Sequential()

model.add(Dense(1, input_shape=(3, ),activation='linear'))

# 학습을 위한 optimizer, 손실함수 loss 정의
model.compile(optimizer=SGD(learning_rate=1e-2), loss='mse')

model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense (Dense)               (None, 1)                 4         
                                                                 
=================================================================
Total params: 4
Trainable params: 4
Non-trainable params: 0
_________________________________________________________________

from datetime import datetime

start_time = datetime.now()

hist = model.fit(x_data, t_data, epochs=1000)

end_time = datetime.now()

print('\nElapsed Time => ', end_time - start_time)

Output exceeds the size limit. Open the full output data in a text editor
Epoch 1/1000
1/1 [==============================] - 0s 477ms/step - loss: 38.2840
Epoch 2/1000
1/1 [==============================] - 0s 8ms/step - loss: 29.0534
Epoch 3/1000
1/1 [==============================] - 0s 10ms/step - loss: 23.9376
Epoch 4/1000
1/1 [==============================] - 0s 9ms/step - loss: 20.8895
Epoch 5/1000
1/1 [==============================] - 0s 8ms/step - loss: 18.8960
Epoch 6/1000
1/1 [==============================] - 0s 10ms/step - loss: 17.4541
Epoch 7/1000
1/1 [==============================] - 0s 7ms/step - loss: 16.3130
Epoch 8/1000
1/1 [==============================] - 0s 7ms/step - loss: 15.3467
Epoch 9/1000
1/1 [==============================] - 0s 9ms/step - loss: 14.4909
Epoch 10/1000
1/1 [==============================] - 0s 7ms/step - loss: 13.7121
Epoch 11/1000
1/1 [==============================] - 0s 7ms/step - loss: 12.9922
Epoch 12/1000
1/1 [==============================] - 0s 7ms/step - loss: 12.3204
Epoch 13/1000
...
Epoch 1000/1000
1/1 [==============================] - 0s 9ms/step - loss: 2.3976e-07

Elapsed Time =>  0:00:21.033955

# 손실함수 출력
import matplotlib.pyplot as plt

plt.title('Loss Trend')
plt.xlabel('epochs')
plt.ylabel('loss')
plt.grid()

plt.plot(hist.history['loss'], label='train loss')
plt.legend(loc='best')

plt.show()

# 예측,  ret_val = 2*x1 -3*x2 + 2*x3
test_data = [ [5, 5, 0], [2, 3, 1], [-1, 0, -1], [10, 5, 2], [4, -1, -2] ]

ret_val = [ 2*data[0] -3*data[1] + 2*data[2]  for data in test_data ]

prediction_val = model.predict(np.array(test_data))

print(prediction_val)
print('==================')
print(ret_val)

[[-5.000464 ]
 [-2.9998715]
 [-3.998861 ]
 [ 8.998674 ]
 [ 7.0006976]]
==================
[-5, -3, -4, 9, 7]

# 모델 입력, 출력, 가중치 확인
print(model.input)
print(model.output)
print(model.weights)

KerasTensor(type_spec=TensorSpec(shape=(None, 3), dtype=tf.float32, name='dense_input'), name='dense_input', description="created by layer 'dense_input'")
KerasTensor(type_spec=TensorSpec(shape=(None, 1), dtype=tf.float32, name=None), name='dense/BiasAdd:0', description="created by layer 'dense'")
[<tf.Variable 'dense/kernel:0' shape=(3, 1) dtype=float32, numpy=
array([[ 1.999865 ],
       [-3.00014  ],
       [ 1.9999071]], dtype=float32)>, <tf.Variable 'dense/bias:0' shape=(1,) dtype=float32, numpy=array([0.00091101], dtype=float32)>]

 

Functional API

 

# Linear Regression 을 keras 이용하여 생성
in_ = Input(shape=(3,))

out_ = Dense(1, activation='linear')(in_)

model = Model(in_, out_)

# 학습을 위한 optimizer, 손실함수 loss 정의
model.compile(optimizer=SGD(learning_rate=1e-2), loss='mse')

model.summary()

Model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input_1 (InputLayer)        [(None, 3)]               0         
                                                                 
 dense_1 (Dense)             (None, 1)                 4         
                                                                 
=================================================================
Total params: 4
Trainable params: 4
Non-trainable params: 0
_________________________________________________________________

from datetime import datetime

start_time = datetime.now()

hist = model.fit(x_data, t_data, epochs=1000)

end_time = datetime.now()

print('\nElapsed Time => ', end_time - start_time)

Output exceeds the size limit. Open the full output data in a text editor
Epoch 1/1000
1/1 [==============================] - 0s 240ms/step - loss: 18.8763
Epoch 2/1000
1/1 [==============================] - 0s 6ms/step - loss: 17.5634
Epoch 3/1000
1/1 [==============================] - 0s 6ms/step - loss: 16.5200
Epoch 4/1000
1/1 [==============================] - 0s 5ms/step - loss: 15.6291
Epoch 5/1000
1/1 [==============================] - 0s 9ms/step - loss: 14.8317
Epoch 6/1000
1/1 [==============================] - 0s 5ms/step - loss: 14.0980
Epoch 7/1000
1/1 [==============================] - 0s 5ms/step - loss: 13.4123
Epoch 8/1000
1/1 [==============================] - 0s 4ms/step - loss: 12.7659
Epoch 9/1000
1/1 [==============================] - 0s 5ms/step - loss: 12.1539
Epoch 10/1000
1/1 [==============================] - 0s 5ms/step - loss: 11.5729
Epoch 11/1000
1/1 [==============================] - 0s 4ms/step - loss: 11.0208
Epoch 12/1000
1/1 [==============================] - 0s 5ms/step - loss: 10.4956
Epoch 13/1000
...
Epoch 1000/1000
1/1 [==============================] - 0s 8ms/step - loss: 1.5905e-11

Elapsed Time =>  0:00:20.754821

# 손실함수 출력
import matplotlib.pyplot as plt

plt.title('Loss Trend')
plt.xlabel('epochs')
plt.ylabel('loss')
plt.grid()

plt.plot(hist.history['loss'], label='train loss')
plt.legend(loc='best')

plt.show()

# 예측,  ret_val = 2*x1 -3*x2 + 2*x3
test_data = [ [5, 5, 0], [2, 3, 1], [-1, 0, -1], [10, 5, 2], [4, -1, -2] ]

ret_val = [ 2*data[0] -3*data[1] + 2*data[2]  for data in test_data ]

prediction_val = model.predict(np.array(test_data))

print(prediction_val)
print('==================')
print(ret_val)

[[-5.0000076]
 [-3.0000012]
 [-3.9999917]
 [ 8.999987 ]
 [ 7.0000057]]
==================
[-5, -3, -4, 9, 7]