Skip to content

representation.md

Latest commit

 

History

History
123 lines (99 loc) · 4.56 KB

representation.md

File metadata and controls

123 lines (99 loc) · 4.56 KB

Representation functions

from codeE.representation import ...

Function to change representation in the crowdsourcing scenario

Examples
... #read some data 
y_obs = ... #classes of shape of (n_samples, n_annotators) not annotation symbol =-1

Change representation to global

from codeE.representation import set_representation
r_obs = set_representation(y_obs,"global")
print("Global representation shape (N,K)= ",r_obs.shape)

Change representation to individual - dense (default of individual)

from codeE.representation import set_representation
y_obs_categorical = set_representation(y_obs, 'onehot') 
print("Individual representation shape (N,T,K)= ",y_obs_categorical.shape)

Change representation to individual - sparse

from codeE.representation import set_representation
y_cat_var, A_idx_var = set_representation(y_obs, 'onehotvar') 
print("Individual sparse representation, variable shape (N,)= ",y_cat_var.shape)
print("one-hot vectors of K-dimensions, K=",y_cat_var[0].shape[1])

From different representation of annotations could be transformed to global

from codeE.representation import annotations2global
r_aux = annotations2global(y_obs)
print(r_aux)
y_obs_categorical = set_representation(y_obs,'onehot') 
r_aux = annotations2global(y_obs_categorical)
print(r_aux)
y_cat_var, _ = set_representation(y_obs,"onehotvar")
r_aux = annotations2global(y_obs)
print(r_aux)

Individual representations could be transformed to global (not the opposite)

Transform representation

codeE.representation.set_representation(y_obs, needed="onehot", no_label = -1, mask_s = 0)

Change representation on crowdsourcing scenario.

Parameters

  • y_obs: array-like of shape (n_samples, n_annotators)
    The multiple annotations observed by the data by every annotators, the class gives by each annotators, if annotator does not annotates there must be a "-1" in the representation.
  • needed: {'onehot','global','onehotvar','variable'}, default="onehot"
    The representation to be seted

'onehot': categorical representation of the annotations by every annotator

'global': number of annotations gives per every label between all the annotators

'onehotvar' or 'variable': categorical representation of variable length, from only anotators that annotate the data.

  • no_label: int, default=-1
    The int value used to find where an annotator does not label some data

  • onehot: int, default=0
    On onehot representation correspond to the symbol to mask when there is no labels for some annotator.

Returns

  • new_annotations: array-like of shape
    The seted representation.

'onehot': (n_samples, n_annotators, n_classes)

'global': (n_samples, n_classes)

'onehotvar' or 'variable': (n_samples,) of arrays of shape (n_annotations(i), n_classes)

  • identity_annotations: same shape as new_annotations
    The identifier of each annotator is returned as well,

Transform annotations to global representation

codeE.representation.annotations2global(obs, no_label=-1)

Change representation from different possible types to global representation.

Parameters

  • obs: array-like of different shapes

(n_samples, n_annotators)

The multiple annotations observed by the data by every annotators, the class gives by each annotators, if annotator does not annotates there must be a "-1" in the representation.

(n_samples, n_annotators, n_classes)

The multiple annotations in a categorical (one hot) representation of the annotations by every annotator (dense setting).

(n_samples,) of arrays of shape (n_annotations(i), n_classes)

The multiple annotations in a categorical (one hot) representation of variable length from only anotators that annotate the data. (sparse setting).

Returns

  • new_annotations: array-like of shape (n_samples, n_classes)
    The global seted representation.

Transform function auxiliar

codeE.representation.list_to_global(list_ann, K)

Change representation from list of variable annotations to global array representation.

Parameters

  • list_ann: list-like of len n_samples
    The multiple annotations observed by the data in a list of variable number of annotations.
  • K: int, n_classes of the data
    The number of the classes of the data

Returns

  • r_obs: array-like of shape (n_samples, n_classes)
    The global representation.