class RNN:
def __init__(self, Wx, Wh, b):
self.params = [Wx, Wh, b]
self.grads = [np.zeros_like(Wx), np.zeros_like(Wh), np.zeros_like(b)]
self.cache = None
def forward(self, x, h_prev)
Wx, Wh, b = self.params
t = np.matmul(h_prev, Wh) + np.matmul(x, Wx) + b
h_next = np.tanh(t)
self.cache = (x, h_prev, h_next)
return h_next
def backward(self, dh_next):
Wx, Wh, b = self.params
x, h_prev, h_next = self.cache
dt = dh_next * (1 - dh_next**2)
db = np.sum(dt, axis=0)
dWh = np.matmul(h_prev.T, dt)
dh_prev = np.matmul(dt, Wh.T)
dWx = np.matmul(x.T, dt)
dx = np.matmul(dt, Wx.T)
self.grads[0][...] = dWx
self.grads[1][...] = dWh
self.grads[2][...] = db
return dx, dh_prev
class TimeRNN:
def __init__(self, Wx, Wh, b, stateful=False):
self.params = [Wx, Wh, b]
self.grads = [np.zeros_like(Wx), np.zeros_like(Wh), np.zeros_like(b)]
self.layers = None # RNN 계층을 리스트로 저장
self.h, self.dh = None, None
self.stateful = stateful # True 시, 은닉상태 유지
def set_state(self, h):
self.h = h
def reset_state(self):
self.h = None
def forward(self, xs):
Wx, Wh, b = self.params
N, T, D = xs.shape # 미니배치, 단어 수, 벡터 차원 수
D, H = Wx.shape
self.layers = []
hs = np.empty((N, T, H), dtype='f')
if not self.stateful or self.h is None: # 처음 호출 시 0으로 초기화
self.h = np.zeros((N,H), dtype='f')
for t in range(T):
layer = RNN(*self.params)
self.h = layer.forward(xs[:, t, :], self.h)
hs[:, t, :] = self.h
self.layers.append(layer)
return hs
def backward(self, dhs):
Wx, Wh, b = self.params
N, T, H = dhs.shape
D, H = Wx.shape
dxs = np.empty((N, T, D), dtype='f')
dh = 0
grads = [0, 0, 0]
for t in reversed(range(T)):
layer = self.layers[t]
dx, dh = layer.backward(dhs[:,t,:] + dh) # 합산된 기울기
dxs[:, t, :] = dx
for i, grad in enumerate(layer.grads):
grads[i] += grad
for i, grad in enumerate(grads):
self.grads[i][...] = grad
self.dh = dh
return dxs
class SimpleRnnlm:
def __init__(self, vocab_size, wordvec_size, hidden_size):
V, D, H = vocab_size, wordvec_size, hidden_size
rn = np.random.randn
# 가중치 초기화
embed_W = (rn(V,D)/100).astype('f')
rnn_Wx = (rn(D,H)/np.sqrt(D)).astype('f')
rnn_Wh = (rn(H,H)/np.sqrt(H)).astype('f')
rnn_b = np.zeros(H).astype('f')
affine_W = (rn(H,V)/np.sqrt(H)).astype('f')
affine_b = np.zeros(V).astype('f')
# 계층 생성
self.layers = [
TimeEmbedding(embed_W),
TimeRNN(rnn_Wx, rnn_Wh, rnn_b, stateful=True),
TimeAffine(affine_W, affine_b)
]
self.loss_layer = TimeSoftmaxWithLoss()
self.rnn_layer = self.layers[1]
# 모든 가중치와 기울기를 리스트에 모은다.
self.params, self.grads = [], []
for layer in self.layers:
self.params += layer.params
self.grads += layer.grads
def forward(self, xs, ts):
for layer in self.layers:
xs = layer.forward(xs)
loss = self.loss_layer.forward(xs, ts)
return loss
def backward(self, dout=1)
dout = self.loss_layer.backward(dout)
for layer in reversed(self.layers):
dout = layer.backward(dout)
return dout
def reset_state(self):
self.rnn_layer.reset_state()
class LSTM:
def __init__(self, Wx, Wh, b):
self.params = [Wx, Wh, b]
self.grads = [np.zeros_like(Wx), np.zeros_like(Wh), np.zeros_like(b)]
self.cache = None
def forward(self, x, h_prev, c_prev):
Wx, Wh, b = self.params
N, H = h_prev.shape
A = np.matmul(x,Wx) + np.matmul(h_prev, Wh) + b
# slice
f = A[:,:H]
g = A[:, H:2*H]
i = A[:, 2*H:3*H]
o = A[:, 3*H:]
f = sigmoid(f)
g = np.tanh(g)
i = sigmoid(f)
o = sigmoid(o)
c_next = f * c_prev + g * i
h_next = o * np.tanh(c_next)
self.cache = (x, h_prev, c_prev, i, f, g, o, c_next)
return h_next, c_next
def backward(self, h_next, c_next):
'''
생략
'''
dA = np.hstack((df, dg, di, do))
'''
생략
'''
참고
밑바닥부터 시작하는 딥러닝2 (사이토 고키)