Utility Functions

All utility functions are importable from NNSOM.utils:

from NNSOM.utils import preminmax, get_cluster_avg, get_perc_cluster

They are also re-exported at the top level of NNSOM.

Data normalization

preminmax

Normalize input data row-wise to the range [-1, 1].

Topology geometry

calculate_positions	Compute the 2-D hexagonal grid positions of SOM neurons.
distances	Compute shortest-path distances between all pairs of SOM neurons.
normalize_position	Normalize neuron positions to the range [-1, 1] along each axis.
spread_positions	Map normalized neuron positions into data space using a basis.
get_hexagon_shape	Return the vertex offsets for drawing a unit hexagon.
get_edge_shape	Return the vertex offsets for drawing the elongated diamond between two neurons.

Coordinate transforms

cart2pol	Convert Cartesian coordinates to polar coordinates.
pol2cart	Convert polar coordinates to Cartesian coordinates.
rotate_xy	Rotate 2-D coordinates by a given angle.

Cluster data extraction

get_cluster_data	For each cluster, extract the corresponding data points.
get_cluster_array	Return an object array whose elements are per-cluster feature values.
get_cluster_avg	Compute the mean feature value for each cluster.
count_classes_in_cluster	Count occurrences of each class in every cluster.
cal_class_cluster_intersect	Calculate the intersection sizes of each class with each neuron cluster.

Class and label analysis

closest_class_cluster	Return the class of the closest data point in each cluster.
majority_class_cluster	Return the majority class label for each cluster.
get_perc_cluster	Return the percentage of a target class in each cluster.
get_color_labels	Generate a colour label per cluster based on majority class membership.
get_edge_widths	Calculate edge line widths per cluster based on class membership fraction.

Error and misclassification analysis

get_ind_misclassified	Return indices of misclassified samples.
get_perc_misclassified	Compute the percentage of misclassified samples in each cluster.
get_conf_indices	Return TP, TN, FP, FN indices for a given class in a binary sense.
get_dominant_class_error_types	Map each cluster's dominant class to its majority error type.

Miscellaneous

flatten	Recursively flatten a nested list of numbers into a single list.
get_global_min_max	Find the global minimum and maximum in a nested list.
create_buttons	Create labelled matplotlib Button widgets in a sidebar.
calculate_button_positions	Calculate evenly-spaced button positions within a figure sidebar.