API reference

Base classes

class deeppy.base.Model[source]

Bases: deeppy.base.ParamMixin

class deeppy.base.ParamMixin[source]

Bases: object

class deeppy.base.PhaseMixin[source]

Bases: object

phase
class deeppy.base.PickleMixin[source]

Bases: object

Fillers

class deeppy.filler.AutoFiller(gain=1.0)[source]

Bases: deeppy.filler.Filler

array(shape)[source]
class deeppy.filler.ConstantFiller(value=0.0)[source]

Bases: deeppy.filler.Filler

array(shape)[source]
class deeppy.filler.CopyFiller(np_array)[source]

Bases: deeppy.filler.Filler

array(shape)[source]
class deeppy.filler.Filler[source]

Bases: object

array(shape)[source]
classmethod from_any(arg)[source]
class deeppy.filler.NormalFiller(mu=0.0, sigma=1.0)[source]

Bases: deeppy.filler.Filler

array(shape)[source]
class deeppy.filler.UniformFiller(low, high)[source]

Bases: deeppy.filler.Filler

array(shape)[source]

Inputs

class deeppy.input.Input(x, batch_size=128)[source]

Bases: object

batches()[source]
classmethod from_any(arg)[source]
shapes
x_shape
class deeppy.input.SupervisedInput(x, y, batch_size=128)[source]

Bases: deeppy.input.Input

batches()[source]
shapes
y_shape

Parameters

class deeppy.parameter.Parameter(fill, name='', learn_rate=1.0, weight_decay=0.0, monitor=False)[source]

Bases: deeppy.base.PickleMixin

array
classmethod from_any(arg)[source]
grad()[source]

Returns a parameter step calculated from the gradient. This differs from grad_array() as the parameter may be shared such that its gradient has multiple sources.

grad_array

Returns the gradient array.

monitor()[source]
penalty()[source]
share()[source]
step(step)[source]

Update the parameter values according to the given step.

class deeppy.parameter.SharedParameter(parent)[source]

Bases: deeppy.parameter.Parameter