Read the latest Real Python tutorials
Project description
Deep-BIAS: Bias In Algorithms, Structural
A toolbox for detecting structural bias in continuous optimization heuristics.
With a deep-learning extension to better evaluate the type of bias and gain insights using explainable AI
Setup
This package requires an R-installation to be present, with the following packages installed:
- PoweR
- AutoSEARCH
- nortest
- data.table
- goftest
- ddst
Detailed setup
- Download and install R from https://cran.r-project.org/
- Download this repository (clone or as zip)
- Create a python virtual env
python -m venv env
- Activate the env (in powershell for example:
env/Scripts/Activate.ps1
) - Install dependencies
pip install -r requirements.txt
- Checkout the
example.py
to start using the BIAS toolbox.
Example
#example of using the BIAS toolbox to test a DE algorithm
from scipy.optimize import differential_evolution
import numpy as np
from BIAS import BIAS, f0, install_r_packages
#run first time to install required R packages
install_r_packages()
bounds = [(0,1), (0, 1), (0, 1), (0, 1), (0, 1)]
#do 30 independent runs (5 dimensions)
samples = []
print("Performing optimization method 30 times of f0.")
for i in np.arange(30):
result = differential_evolution(f0, bounds, maxiter=100)
samples.append(result.x)
samples = np.array(samples)
test = BIAS()
print(test.predict(samples, show_figure=True))
y, preds = test.predict_deep(samples)
test.explain(samples, preds, filename="explanation.png")
Additional files
Note: The code for generating the RF used to predict the type of bias is included, but the full RF is not. These can be found on zenodo: https://doi.org/10.6084/m9.figshare.16546041. The RF models will be downloaded automatically the first time the predict function requires them.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
struct-bias-1.1.0.tar.gz
(15.7 kB
view hashes)
Built Distribution
Close
Hashes for struct_bias-1.1.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | b8aabbf64d73db450c9c9a391432a3acdd93f1d38b5909362995058fdd00cda2 |
|
MD5 | 6199a8dd399168fe42ff8addaf709e4b |
|
BLAKE2b-256 | 4382de800758c15cafb0b03056dd7433db014fbc3f3bf6ffff68398587913790 |