Building off of the simulations implemented in #17, we wanted to see how things would change as we either add spurious edges to or remove true edges from the networks used to generate the regression coefficients. Our results suggest that the method is much more sensitive to adding spurious edges than to removing true edges, which is interesting.
Changes in this PR:
Add code to simulate_network.py to add/remove edges
Add analysis notebook edge_analysis.ipynb
Modify netreg_benchmark.py to change add/remove percentage and write hyperparameter choices to a file
Building off of the simulations implemented in #17, we wanted to see how things would change as we either add spurious edges to or remove true edges from the networks used to generate the regression coefficients. Our results suggest that the method is much more sensitive to adding spurious edges than to removing true edges, which is interesting.
Changes in this PR:
simulate_network.pyto add/remove edgesedge_analysis.ipynbnetreg_benchmark.pyto change add/remove percentage and write hyperparameter choices to a filesimulation_analysis.ipynb