Model selection - Githubissues

pathpy / pathpyG

GPU-accelerated Next-Generation Network Analytics and Graph Learning for Time Series Data on Complex Networks.

https://www.pathpy.net

GNU Affero General Public License v3.0

33 stars 3 forks source link

Model selection #173

Closed VinsRR closed 2 months ago

VinsRR commented 6 months ago

This branch proposes the following modifications:

Implementation of model selection for MultiOrderModel (for walk data only)
tests for the modules related to model selection in MultiOrderModel
functions compute_weighted_outdegree and compute_transition_probabilities (necessary to perform model selection. These functions are located in MultiOrderModel.py
fixing weights for higher-order networks (thus making the branch fix-hon-weights obsolete) #157

The modifications in this branch also relate to the following topics:

Edge weight representation #155 -- workarounds have been developed to relate edge frequencies to path frequencies (as, currently, path frequencies are saved into edge frequencies). See all the frequencies variables in MultiOrderModel
Zeroth order layer: workarounds have been developed to perform the model selection without needing the zeroth order in MultiOrderModel. The question remains whether we want to implement one or not #172

IngoScholtes commented 2 months ago

All tests succeeded, I will merge this PR.