multiview_generator.gaussian_classes

gaussian_classes

class MultiViewGaussianSubProblemsGenerator(random_state=42, n_samples=100, n_classes=4, n_views=4, error_matrix=None, n_features=3, class_weights=1.0, redundancy=0.05, complementarity=0.05, complementarity_level=3, mutual_error=0.01, name='generated_dataset', config_file=None, sub_problem_type='base', sub_problem_configurations=None, sub_problem_generators='StumpsGenerator', random_vertices=False, min_rndm_val=-1, max_rndm_val=1, **kwargs)

This engine generates one monoview sub-problem for each view with independant data. If then switch descriptions between the samples to create error and difficulty in the dataset

Parameters:

  • random_state – The random state or seed.

  • n_samples – The number of samples that the dataset will contain

  • n_classes – The number of classes in which the samples will be labelled

  • n_views – The number of views describing the samples

  • error_matrix – The error matrix giving in row i column j the error of the Bayes classifier on Class i for View j

  • n_features – The number of features describing the samples for each view (can specify an int or array-like of length n_views)

  • class_weights – The proportion of the dataset that will be labelled in each class. Must specify an array-like of size n_classes ([0.1,0.45,0.45] will output a dataset with with 10% of the samples in the first class and 45% in the two others.)

  • redundancy – The proportion of the samples that will be well-decribed by all the views.

# :param complementarity: The proportion of samples that will be well-decribed only by some views :param complementarity_level: The number of views that will have a bad description of the complementray samples :param mutual_error: The proportion of samples that will be mis-described by all the views :param name: The name of the dataset (will be used to name the file) :param config_file: The path to the yaml config file. If provided, the config fil entries will overwrite the one passed as arguments.

random_vertices
sub_problem_generators
generate_multi_view_dataset()

This is the main method. It will generate a multiview dataset according to the configuration. To do so,

  • it generates the labels of the multiview dataset,

  • then it assigns all the subsets of samples (redundant, …)

  • finally, for each view it generates a monoview dataset according to the configuration

Returns:

view_data a list containing the views np.ndarrays and y, the label array.

assign_mutual_error()

Method assigning the mis-describing views to the mutual error samples.

assign_complementarity()

Method assigning mis-described and well-described views to build complementary samples

assign_redundancy()

Method assigning the well-describing views to the redundant samples.

get_distance()

Method that records the distance of each description to the ideal decision limit, will be used later to quantify more precisely the quality of a description.