Learning outside the black-box: at the pursuit of interpretable models
Project description
Symbolic Pursuit
Github for the NIPS 2020 paper "Learning outside the black-box: at the pursuit of interpretable models"
Installation
The library can be installed from PyPI using
$ pip install symbolic_pursuit
or from source, using
$ pip install .
Example Usage
To build a symbolic regressor for a given dataset and a given model (or a given model type), the following command can be used :
python3 build_interpreter.py [-h] [--dataset DATASET] [--test_ratio TEST_RATIO]
[--model MODEL] [--model_type MODEL_TYPE]
[--verbosity VERBOSITY] [--loss_tol LOSS_TOL]
[--ratio_tol RATIO_TOL] [--maxiter MAXITER]
[--eps EPS] [--random_seed RANDOM_SEED]
For example, if one would like to train a MLP one the wine-quality-red dataset and then fit a symbolic regressor with random seed 27, one can use the command
python3 build_interpreter --dataset wine-quality-red --model_type MLP --random_seed 27
For more details on how to use the module in general, see the 3 enclosed notebooks.
1. Building a Symbolic Regressor 2. Symbolic Pursuit vs LIME 3. Synthetic experiments with Symbolic Pursuit
:hammer: Tests
Install the testing dependencies using
pip install .[testing]
The tests can be executed using
pytest -vsx
References
In our experiments, we used implementations of LIME, SHAP and pysymbolic
Citing
If you use this code, please cite the associated paper:
@article{https://doi.org/10.48550/arxiv.2011.08596,
doi = {10.48550/ARXIV.2011.08596},
url = {https://arxiv.org/abs/2011.08596},
author = {Crabbé, Jonathan and Zhang, Yao and Zame, William and van der Schaar, Mihaela},
keywords = {Machine Learning (cs.LG), Artificial Intelligence (cs.AI), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Learning outside the Black-Box: The pursuit of interpretable models},
publisher = {NeurIPS 2020},
year = {2020},
}
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 Distributions
Built Distributions
Hashes for symbolic_pursuit-0.0.1-py3-none-macosx_10_14_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 1708bd22d9462be82cffaf86fc35cf93e40e7a44ae7103fe6303ece9aaed062d |
|
MD5 | 6046ba7cc9d4b45a8eb4786e7ab87838 |
|
BLAKE2b-256 | 0c5a3c7b91246e5909990423c987ad67d0ead32b4159cbf77f0635d955538c0f |
Hashes for symbolic_pursuit-0.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3e548a738d669d0f7563512c49ae59fb08a9d6dadb706a759cc1d3a248f8e95a |
|
MD5 | c149841aa7b96723c97b9155035525e1 |
|
BLAKE2b-256 | b57bab8a03b70834956c95fcaf8839e19dd3bf5ddbbbf5864fcce40b8f5a58dc |