View Code? Open in Web Editor
NEW
Analysis of (group) fairness measures. How fair can you get under class imbalance. Representation bias and stereotypical bias. Histogram visualizations and analyses.
Python 0.40%
Jupyter Notebook 99.60%
analysis-of-fairness-measures's People
Watchers
analysis-of-fairness-measures's Issues
y = count, x = FM on 0.01, 0.1, 0.5 ... for FM: MinR. GR, SB ... (test e.g equality of % of hires of group1 and group2)
play with the number of bins
normalize y axis to probabilities tricky
find where NULLs occur while computing FMs, count them, keep them out of hists (plot aside) very tricky
Move functions to a separate utils.py
and use importing it.
Come up with more tests.
Add requirements.txt
.
For a given n
generate all possible binary group confusion matrices, e.g for n = 10
TP_m
TN_m
FP_m
FN_m
TP_f
TNf_
FP_f
FN_f
10
0
0
0
0
0
0
0
9
1
0
0
0
0
0
0
...
...
...
...
...
...
...
...
Write the goal of the project, mention original repo, add environment set up tutorial.
Check if the the data is generated correctly. GR measure may be used.
Precess dataset with appropriate names.
Create functions for computing, make plots (MinR, GR, SB ...).
Fix var names, last plots.
Populate calculations
dir with computation for smaller n
as an example.
Come up with a way of storing and accessing the main data remotely.