Class InputOutputDeepArchitecture
source code
An InputOutputDeepArchitecture (IODA) is a specialization of the DNN,
where the layers are divided into three categories : the input layers,
the link layer and the output layers. It has been specifically designed
for cases where both the input and the output spaces are high-dimensional.
The input and output layers are pretrained on the training example
(x) and the training labels (y),
respectively, using a Stacked AutoEncoder strategy, as for DNNs.
The link layer can optionally be pretrained, using as input and output data
the hidden representations of the deepmost input and output autoencoders,
respectively.
|
|
Inherited from PretrainedMLP :
prepareParams ,
pretrainInputAutoEncoders ,
pretrainOutputAutoEncoders ,
train
Inherited from MultiLayerPerceptron :
checkBadmoveHook ,
checkBatchHook ,
checkEpochHook ,
checkLearningParameters ,
defaultLearningParameters ,
finetune ,
getGeometry ,
getParameters ,
initBadmoveHook ,
initBatchHook ,
initEpochHook ,
setParameters
Inherited from module.Sequential :
prepareGeometry ,
prepareOutput
Inherited from module.Container :
add
Inherited from module.Module :
criterionFunction ,
forward ,
forwardFunction ,
holdFunction ,
linkInputs ,
linkModule ,
prepare ,
prepareBackup ,
restoreFunction ,
save ,
trainFunction
|
__init__(self,
nUnitsInput,
nUnitsOutput,
outputActivation=<class crino.module.Sigmoid at 0x2b85740f5db8>)
(Constructor)
| source code
|
Constructs a new InputOutputDeepArchitecture.
- Parameters:
nUnitsInput (int list) - The sizes of the (input and hidden) representations on the input side.
nUnitsOutput (int list) - The sizes of the (hidden and output) representations on the output side.
outputActivation (class derived from Activation) - The type of activation for the output layer.
- Overrides:
module.Module.__init__
Attention:
outputActivation parameter is not an instance but a class.
|