def __init__(self, start_shape, stop_shape, depth=2, conv_part=Conv1dAC):
_Net.__init__(self, start_shape, stop_shape, depth)
self.conv_part = conv_part(start_shape=start_shape,
stop_shape=stop_shape,
depth=depth,
bias_prob=0.3)
self.lin_part = LinearAC(start_shape=None,
stop_shape=(10, ),
depth=depth,
bias_prob=0.3)
self.change_num = IntParam(name="")
def __init__(
self,
start_shape,
stop_shape,
min_features=10,
max_features=128,
depth=2,
bias_prob=0.0,
):
_Net.__init__(self, start_shape, stop_shape, depth)
self.layers = [Linear(bias=False)]
self.limits = (min_features, max_features)
if bias_prob > 0:
self.layers[0].bias.randomize(true_prob=bias_prob)
self.layers[0].out_features.randomize(limits=self.limits)
def __call__(self, num_nets, startnum=1):
t = self.layers[0].bias
if t.is_random and t.true_prob > 0:
if "Only" in self.name:
self.name = self.name.replace("Only", "Bias")
elif "Bias" not in self.name:
self.name = self.name + "Bias"
return _Net.__call__(self, num_nets, startnum)
def __call__(self, num_nets, startnum=1):
self.name = "{}".format(
self.__class__.__name__.replace("AC",
self.ac.val.__class__.__name__))
return _Net.__call__(self, num_nets, startnum)
def __init__(self, start_shape, stop_shape, depth, convclass, bias_prob=0):
_Net.__init__(self, start_shape, stop_shape, depth)
self.layers = [convclass()]
self.layers[0].out_channels.randomize(limits=(self.stop_shape[0], 64))
if bias_prob > 0:
self.layers[0].bias.randomize(true_prob=bias_prob)
def __call__(self, num_nets, startnum=1):
self.name = "Linear{}".format(self.ac.val.__class__.__name__)
return _Net.__call__(self, num_nets, startnum)
def __init__(self, start_shape, stop_shape, depth=3):
_Net.__init__(self, start_shape, stop_shape, depth)
self.layers.append(Conv2dAC(start_shape, [10, 1, 1], depth))
self.layers.append(Conv2dThenLinear(start_shape, [10, 1, 1], depth))
self.change_num = IntParam("", default=1)