Metadata-Version: 2.1
Name: valda
Version: 0.1.10
Summary: A Data Valuation Package for Machine Learning
Author-email: Yangfeng Ji <yangfeng@virginia.edu>, Stephanie Schoch <sns2gr@virginia.edu>
Project-URL: Homepage, https://uvanlp.org/valda
Project-URL: Bug Tracker, https://github.com/uvanlp/valda/issues
Keywords: data valuation
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.6
Description-Content-Type: text/markdown
License-File: LICENSE.rst

# Valda

## Introduction

Valda is a Python package for data valuation in machine learning. If you are interested in 

- analyzing the contribution of individual training examples to the final classification performance, or 
- identifying some noisy examples in the training set, 

you may be interested in the functions provided by this package.


The current version supports five different data valuation methods. It supports all the classifiers from Sklearn for valuation, and also user-defined classifier using PyTorch. 

- Leave-one-out (LOO), 
- Data Shapley with the TMC algorithm (TMC-Shapley) from [Ghorbani and Zou (2019)](https://proceedings.mlr.press/v97/ghorbani19c.html), 
- Beta Shapley from [Kwon and Zou (2022)](https://arxiv.org/abs/2110.14049)
- Class-wise Shapley (CS-Shapley) from [Schoch et al. (2022)](https://arxiv.org/abs/2211.06800)
- Influence Function (IF) from [Koh and Liang (2017)](https://arxiv.org/abs/1703.04730)
  - IF only works with the classifiers built with PyTorch, because it requires gradient computation.
  - v0.1.8 only support the first-order gradient computation, and we will add the second-order computation soon.




## Tutorial

Please checkout a simple tutorial on [Google Colab](https://colab.research.google.com/drive/1agsMNqZan-3RnJLQtBGATRHHWYMe7C9H?usp=sharing), for how to use this package. 
