def feed_backwards(self, x: np.ndarray) -> np.ndarray:
v = np.zeros((1, 28 * 28))
v[:, 0:10] = x
x = homogenize_vector(v)
for m in self.R:
x[:, :-1] = self.ai(np.clip(x[:, :-1], 0.000000001, 0.999999999))
x = np.matmul(x, m)
return x[:, :-1]
def feed_backwards(self, x: np.ndarray) -> np.ndarray:
x = homogenize_vector(x)
for m in self.R:
x[:, :-1] = self.ai(np.clip(x[:, :-1], 0.000000001, 0.999999999))
x = np.matmul(x, m)
return x[:, :-1]
def feed_forward(self, x: np.ndarray) -> np.ndarray:
x = homogenize_vector(x)
for m in self.M:
x = np.matmul(x, m)
x[:, :-1] = self.a(np.clip(x[:, :-1], -500, 500))
return x[:, :-1]
def feed_backwards(self, x: np.ndarray) -> np.ndarray:
x = homogenize_vector(self.ai(np.clip(x, 0.000000001, 0.999999999)))
x = np.matmul(x, self.R)
return x[:, :-1]
def feed_backwards(self, x: np.ndarray) -> np.ndarray:
x = homogenize_vector(x)
x = np.matmul(x, self.R)
return x[:, :-1]
def feed_forward(self, x: np.ndarray) -> np.ndarray:
x = homogenize_vector(x)
x = np.matmul(x, self.M)
return x[:, :-1]
