Skip to content

Granular class (combination) selection in multiclassification cases.

Artur Monsch requested to merge dev_pandas_overlay into master

Implementation of more granular selection of a class (combination) in case of multiclassification.

Example: Neural network with three output nodes, the following selections are possible

  • eval_nodes=0: Only the consideration of one class/node
  • eval_nodes='all': All classes/nodes are summed up
  • eval_nodes=[0, 1, (0, 1)]: Classes 0 and 1 are considered individually and summed up
  • eval_nodes=[0, (0, 1), (1, 2), 'all']: Combination of all cases above

Goal: Better understanding of neural network learning with respect to specific learning goals: Classes (combinations).

Merge request reports