def __init__(self, initializer=None, shape=None, name=None):
if initializer is None:
initializer = constant.NaN()
elif numpy.isscalar(initializer):
initializer = constant.Constant(initializer)
if shape is None:
if isinstance(initializer, (numpy.ndarray, cuda.ndarray)):
# parameter initialized by the initial array
super(Parameter, self).__init__(initializer, name=name)
else:
# uninitialized parameter
super(Parameter, self).__init__(name=name)
self.initializer = initializer
dtype = getattr(initializer, 'dtype', numpy.float32)
self._grad_initializer = constant.NaN(dtype)
else:
# parameter initialized with a given shape
if isinstance(initializer, (numpy.ndarray, cuda.ndarray)):
xp = cuda.get_array_module(initializer)
initializer = constant.Constant(initializer)
else:
xp = numpy
data = initializers.generate_array(initializer, shape, xp)
grad = xp.full_like(data, numpy.nan)
super(Parameter, self).__init__(data, name=name, grad=grad)
self.update_rule = None
def setUp(self):
self.in_size = 2
self.out_size = 3
self.pool_size = 4
if self.initializer == 'Initializer':
self.initialW = constant.Constant(1.0)
self.initial_bias = constant.Constant(2.0)
elif self.initializer == 'scalar':
self.initialW = 1.0
self.initial_bias = 2.0
elif self.initializer == 'ndarray':
self.initialW = numpy.random.uniform(
-1, 1, (self.out_size, self.pool_size, self.in_size)).astype(
self.dtype)
self.initial_bias = numpy.random.uniform(
-1, 1, (self.out_size, self.pool_size)).astype(self.dtype)
elif self.initializer == 'callable':
def callable_initialW(array):
assert array.dtype == self.dtype
assert array.shape == (self.out_size, self.pool_size,
self.in_size)
array.fill(1.0)
self.initialW = callable_initialW
def callable_initial_bias(array):
assert array.dtype == self.dtype
assert array.shape == (self.out_size, self.pool_size)
array.fill(2.0)
self.initial_bias = callable_initial_bias
else:
raise ValueError('invalid parameter')
with chainer.using_config('dtype', self.dtype):
self.link = links.Maxout(self.in_size,
self.out_size,
self.pool_size,
initialW=self.initialW,
initial_bias=self.initial_bias)
