HALF.Strategies#

HALF.Strategies.wrap_distil_util#

Strategies adapted from cure-lab#

CoreGCN#

Graph convolutional network sampling strategies (UncertainGCN, CoreGCN, CVPR’21) Sequential Graph Convolutional Network for Active Learning inspired by the implementation from cure-lab https://github.com/cure-lab and by the main the paper https://github.com/razvancaramalau/Sequential-GCN-for-Active-Learning/blob/master/selection_methods.py

param labeled_dataset:: labelled dataset
type labeled_dataset:: Dataset
param unlabeled_dataset:: unlabelled dataset
type unlabeled_dataset:: Dataset
param net:: network model to use
type net:: Module
param nclasses:: number of classes for the model
type nclasses:: int
param args:: can contain the following parameters: - n_hidden: number of hidden units from the GCN used during the select method - dropout: dropout rate of the GCN. Experiments tend to show not to go above 0.3 - subset: we do not select on all unlabeled data but rather on a subset of this size within it - lambda_loss: lambda, weighting factor between the labelled and unlabelled cross-entropy - epoch_gcn number of epochs for the internal GCN model - lr_gcn GCN learning rate - wdecay: weight decay for the GCN default values are constant at the top of the file
type args:: dict

HALF.Strategies.sequence_gcn.CoreGCN.__delattr__(self, name, /)#: Implement delattr(self, name).

HALF.Strategies.sequence_gcn.CoreGCN.__dir__(self, /)#: Default dir() implementation.

HALF.Strategies.sequence_gcn.CoreGCN.__eq__(self, value, /)#: Return self==value.

HALF.Strategies.sequence_gcn.CoreGCN.__format__(self, format_spec, /)#: Default object formatter.

HALF.Strategies.sequence_gcn.CoreGCN.__ge__(self, value, /)#: Return self>=value.

HALF.Strategies.sequence_gcn.CoreGCN.__getattribute__(self, name, /)#: Return getattr(self, name).

HALF.Strategies.sequence_gcn.CoreGCN.__gt__(self, value, /)#: Return self>value.

HALF.Strategies.sequence_gcn.CoreGCN.__hash__(self, /)#: Return hash(self).

HALF.Strategies.sequence_gcn.CoreGCN.__init_subclass__()#

This method is called when a class is subclassed.

The default implementation does nothing. It may be overridden to extend subclasses.

HALF.Strategies.sequence_gcn.CoreGCN.__le__(self, value, /)#: Return self<=value.

HALF.Strategies.sequence_gcn.CoreGCN.__lt__(self, value, /)#: Return self<value.

HALF.Strategies.sequence_gcn.CoreGCN.__ne__(self, value, /)#: Return self!=value.

HALF.Strategies.sequence_gcn.CoreGCN.__new__(*args, **kwargs)#: Create and return a new object. See help(type) for accurate signature.

HALF.Strategies.sequence_gcn.CoreGCN.__reduce__(self, /)#: Helper for pickle.

HALF.Strategies.sequence_gcn.CoreGCN.__reduce_ex__(self, protocol, /)#: Helper for pickle.

HALF.Strategies.sequence_gcn.CoreGCN.__repr__(self, /)#: Return repr(self).

HALF.Strategies.sequence_gcn.CoreGCN.__setattr__(self, name, value, /)#: Implement setattr(self, name, value).

HALF.Strategies.sequence_gcn.CoreGCN.__sizeof__(self, /)#: Size of object in memory, in bytes.

HALF.Strategies.sequence_gcn.CoreGCN.__str__(self, /)#: Return str(self).

HALF.Strategies.sequence_gcn.CoreGCN.__subclasshook__()#

Abstract classes can override this to customize issubclass().

This is invoked early on by abc.ABCMeta.__subclasscheck__(). It should return True, False or NotImplemented. If it returns NotImplemented, the normal algorithm is used. Otherwise, it overrides the normal algorithm (and the outcome is cached).

UncertainGCN#

HALF.Strategies.sequence_gcn.UncertainGCN.__delattr__(self, name, /)#: Implement delattr(self, name).

HALF.Strategies.sequence_gcn.UncertainGCN.__dir__(self, /)#: Default dir() implementation.

HALF.Strategies.sequence_gcn.UncertainGCN.__eq__(self, value, /)#: Return self==value.

HALF.Strategies.sequence_gcn.UncertainGCN.__format__(self, format_spec, /)#: Default object formatter.

HALF.Strategies.sequence_gcn.UncertainGCN.__ge__(self, value, /)#: Return self>=value.

HALF.Strategies.sequence_gcn.UncertainGCN.__getattribute__(self, name, /)#: Return getattr(self, name).

HALF.Strategies.sequence_gcn.UncertainGCN.__gt__(self, value, /)#: Return self>value.

HALF.Strategies.sequence_gcn.UncertainGCN.__hash__(self, /)#: Return hash(self).

HALF.Strategies.sequence_gcn.UncertainGCN.__init_subclass__()#