예시 데이터 생성
def temporalize(X, y, lookback):
output_X = []
output_y = []
for i in range(len(X)-lookback-1): # i=0 ~ 4, when lookback=3
t = []
for j in range(1, lookback+1): # j=1 ~ 3, when lookback=3
t.append(X[[i+j+1], :])
output_X.append(t) # i=0, t.append X[2] ~ X[4]
# i=1, t.append X[3] ~ X[5]
output_y.append(y[i+lookback+1]) # i=0, y = y[4]
# i=1, y = y[5]
return output_X, output_ytimeseries = np.array([[0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9],
[0.1**3, 0.2**3, 0.3**3, 0.4**3, 0.5**3, 0.6**3, 0.7**3, 0.8**3,
0.9**3]]).transpose()
timesteps = timeseries.shape[0]
n_features = timeseries.shape[1]
timeseries # shape=(9, 2)
'''
array([[0.1 , 0.001],
[0.2 , 0.008],
[0.3 , 0.027],
[0.4 , 0.064],
[0.5 , 0.125],
[0.6 , 0.216],
[0.7 , 0.343],
[0.8 , 0.512],
[0.9 , 0.729]])
'''timesteps = 3
X, y = temporalize(X=timeseries, y=np.zeros(len(timeseries)), lookback = timesteps)
n_features = 2
X = np.array(X)
X = X.reshape(X.shape[0], timesteps, n_features)
# X.shape = (5, 3, 2)
'''
X =
[[[0.3 0.027]
[0.4 0.064]
[0.5 0.125]]
[[0.4 0.064]
[0.5 0.125]
[0.6 0.216]]
[[0.5 0.125]
[0.6 0.216]
[0.7 0.343]]
[[0.6 0.216]
[0.7 0.343]
[0.8 0.512]]
[[0.7 0.343]
[0.8 0.512]
[0.9 0.729]]]
'''
'''
np.zeros(len(timeseries)) =
[0. 0. 0. 0. 0. 0. 0. 0. 0.]
'''
LSTM Autoencoder 구조 이해하기
model = keras.models.Sequential()
model.add(keras.layers.LSTM(128, activation='relu', input_shape=(timesteps, n_features),
return_sequences=True))
model.add(keras.layers.LSTM(64, activation='relu', return_sequences=False))
model.add(keras.layers.RepeatVector(timesteps))
model.add(keras.layers.LSTM(64, activation='relu', return_sequences=True))
model.add(keras.layers.LSTM(128, activation='relu', return_sequences=True))
model.add(keras.layers.TimeDistributed(keras.layers.Dense(n_features)))
model.compile(optimizer='adam', loss='mse')
model.summary()
'''
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
lstm_4 (LSTM) (None, 3, 128) 67072
lstm_5 (LSTM) (None, 64) 49408
repeat_vector_1 (RepeatVect (None, 3, 64) 0
or)
lstm_6 (LSTM) (None, 3, 64) 33024
lstm_7 (LSTM) (None, 3, 128) 98816
time_distributed_1 (TimeDis (None, 3, 2) 258
tributed)
=================================================================
Total params: 248,578
Trainable params: 248,578
Non-trainable params: 0
_________________________________________________________________
'''# fit the model
model.fit(X, X, epochs=300, batch_size=5, verbose=0)
# reconstruction
yhat = model.predict(X, verbose=0)
print('---Predicted---')
print(np.round(yhat, 3))
print('---Actual---')
print(np.round(X, 3))
'''
---Predicted---
[[[0.321 0.04 ]
[0.413 0.064]
[0.497 0.124]]
[[0.39 0.068]
[0.494 0.122]
[0.595 0.214]]
[[0.492 0.12 ]
[0.595 0.215]
[0.702 0.341]]
[[0.603 0.209]
[0.698 0.348]
[0.798 0.509]]
[[0.697 0.343]
[0.797 0.508]
[0.894 0.725]]]
---Actual---
[[[0.3 0.027]
[0.4 0.064]
[0.5 0.125]]
[[0.4 0.064]
[0.5 0.125]
[0.6 0.216]]
[[0.5 0.125]
[0.6 0.216]
[0.7 0.343]]
[[0.6 0.216]
[0.7 0.343]
[0.8 0.512]]
[[0.7 0.343]
[0.8 0.512]
[0.9 0.729]]]
'''
위 다이어그램은 한 데이터 샘플에 대한 데이터 흐름입니다. ([[0.4, 0.5, 0.6], [0.064, 0.125, 0.216]])
RepeatVector(timesteps)를 사용해 벡터를 timesteps 만큼 복제합니다.
'핸즈온 머신러닝 2판' 카테고리의 다른 글
| 18장 강화 학습 (0) | 2022.02.07 |
|---|---|
| 이진 오토인코더를 이용한 해싱 (0) | 2022.02.06 |
| 17장 오토인코더와 GAN을 사용한 표현 학습과 생성적 학습 (2) | 2022.02.03 |
| 16장 RNN과 어텐션을 사용한 자연어 처리 (1) | 2022.01.26 |
| 15장 RNN과 CNN을 사용해 시퀀스 처리하기 (0) | 2022.01.21 |