def orthogonal(shape, scale=1.1, name=None):
flat_shape = (shape[0], np.prod(shape[1:]))
a = np.random.normal(0.0, 1.0, flat_shape)
u, _, v = np.linalg.svd(a, full_matrices=False)
q = u if u.shape == flat_shape else v # pick the one with the correct shape
q = q.reshape(shape)
return sharedX(scale * q[:shape[0], :shape[1]], name=name)
def identity(shape, scale=1.0, name=None):
if len(shape) != 2 or shape[0] != shape[1]:
raise Exception("Identity matrix initialization can only be used for 2D square matrices")
else:
return sharedX(scale * np.identity(shape[0]), name=name)
def normal(shape, scale=0.05, name=None):
return sharedX(np.random.randn(*shape) * scale, name=name)
def uniform(shape, scale=0.05, name=None):
return sharedX(np.random.uniform(low=-scale, high=scale, size=shape), name=name)
