DNN (Deep Neural Network)

DNN (Deep Neural Network)

Keras 기본

선형 신경망

model = Sequential()
model.add(Dense(1, input_dim=2))
model.add(Activation('sigmoid'))
model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_2 (Dense)              (None, 1)                 3         
_________________________________________________________________
activation_2 (Activation)    (None, 1)                 0         
=================================================================
Total params: 3
Trainable params: 3
Non-trainable params: 0
_________________________________________________________________

신경망 하나만 사용한 것.

입력은 2개(input dimension), 출력 1개

-> 파라미터 3개 (w1, w2, b)

model = Sequential()
model.add(Dense(5, input_dim=2))    #입력층 5*2+5=15개 (학습 파라미터 수)
model.add(Activation('sigmoid'))
model.add(Dense(1))                 #출력층 
model.add(Activation('sigmoid'))
model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_3 (Dense)              (None, 5)                 15        
_________________________________________________________________
activation_3 (Activation)    (None, 5)                 0         
_________________________________________________________________
dense_4 (Dense)              (None, 1)                 6         
_________________________________________________________________
activation_4 (Activation)    (None, 1)                 0         
=================================================================
Total params: 21
Trainable params: 21
Non-trainable params: 0
_________________________________________________________________

학습 파라미터 수: 입력층 5*(2+1)=15개 + 출력층 6개 = 21개

 

model.compile(loss='binary_crossentropy', optimizer='adam') #모델 컴파일
model.fit(X,y,batch_size=4, epochs=2000, verbose=0)
print(model.predict_proba(X)) #4x1 matrix

p = model.predict(X)
print(p > 0.5)
print((p > 0.5)*1) #binary로 표현