1. 16 Nov, 2021 1 commit
  2. 15 Nov, 2021 4 commits
  3. 13 Sep, 2021 1 commit
    • Noric Couderc's avatar
      Print raw features without aggregation · 98262034
      Noric Couderc authored
      We used to run the benchmark many times are aggregate the values of
      counters (median) before. Now when we print the features, we don't do
      that anymore. We print the raw data.
      
      Doing this required to add an "iteration" number for each specific run.
      98262034
  4. 22 Jun, 2021 4 commits
  5. 14 Jun, 2021 1 commit
  6. 02 Jun, 2021 3 commits
  7. 31 May, 2021 2 commits
  8. 25 May, 2021 1 commit
  9. 19 May, 2021 4 commits
    • Noric Couderc's avatar
      We keep samples where the ratio is 1. · 5fd9e116
      Noric Couderc authored
      In the previous setup, samples where the best datastructure was the
      current considered (the winning samples, in a way), were tossed away.
      I think keeping them helps the classifier notice what works.
      
      This improved the classification accuracy with only software features.
      5fd9e116
    • Noric Couderc's avatar
      Refactoring: extracted variables · c71f5d2d
      Noric Couderc authored
      Assigned sys.argv elements to variables with their own names.
      
      That's mostly to make running code in the REPL easier.
      c71f5d2d
    • Noric Couderc's avatar
      Dropped usage of Lower_score in classification · adc1438a
      Noric Couderc authored
      We used to use the "Lowest score" for classification (which is the
      score minus the error). We don't use it anymore, because we will drop
      the elements that don't have a significant improvement over the
      baseline.
      
      There is probably a way to take the error of the measurements into
      account but I'm thinking this is not a sound way to do it.
      adc1438a
    • Noric Couderc's avatar
      Generalized function train_classifiers · cc7c632a
      Noric Couderc authored
      We pass the column for the labels, the features, and the grouping
      column as arguments to the function now.
      
      This was done for some experiments with using only the software
      features for classification.
      cc7c632a
  10. 18 May, 2021 2 commits
  11. 13 May, 2021 6 commits
  12. 12 May, 2021 1 commit
  13. 07 May, 2021 2 commits
  14. 05 May, 2021 4 commits
  15. 04 May, 2021 1 commit
  16. 13 Apr, 2021 3 commits