Skip to content

Latest commit

 

History

History
executable file
·
11 lines (6 loc) · 1.12 KB

README.md

File metadata and controls

executable file
·
11 lines (6 loc) · 1.12 KB

Reproducing Results

To reproduce results given in the paper, follow the below steps:

  1. Generate Models: Run any of the .sh files in this directory to create model and data dumps for a data setting. The files have been named run_D_mMnN.sh where M is the manifold dimension and N is the embedding dimension, and D is an identifier for the dataset type (cs: Concentric Spheres, sw: Intertwined Swiss Rolls, ws: Separated Spheres).

  2. Decision Region & Heatmap Plots: Navigate to ppr_decreg_and_heatmaps.py. Provide the path to the distance learner and standard classifier dumps, and an identifier for the plot file names and run the script.

  3. Out-of-domain Confidence Plot: Navigate to ppr_confidence.py. Provide the path to the distance learner and standard classifier dumps, and an identifier for the plot file names and run the script.

  4. Adversarial Robustness Plot: Make sure you have run run_cs_m50n500.sh and run_cs_m25n500.sh. Navigate to adv_robustness.py. Provide a path to the locations of adversarial performance dumps (.json files created when you run the bash scripts), and run the script.