Skip to content

Latest commit

 

History

History
245 lines (190 loc) · 50.7 KB

README.md

File metadata and controls

245 lines (190 loc) · 50.7 KB

Awesome Data Valuation

data market problem

💱 A curated list of data valuation (DV) to design your next data marketplace. DV aims to understand the value of a data point for a given machine learning task and is an essential primitive in the design of data marketplaces and explainable AI.

Legend

💻 Code available

🎥 Talk / Slides

Contents

Table of Contents
1. What is your data worth? (DV Algorithms)
    1.1. Shapley Value & Cooperative Game Theory
        1.1.1. Efficient algorithms
        1.1.2. Benchmarks, Criticism & Relaxations
    1.2. Influence functions & LOO
    1.3. Reinforcement Learning
    1.4. Deep Neural Networks
    1.5. Out-of-Bag score
    1.6. Task Agnostic
2. Benchmarks
3. Libraries
    3.1. Surveys
4. Designing data marketplaces
    4.1. Data market system designs
    4.2. Automatic data compliance
    4.3. Data valuation applications
5. Data markets and society
    5.1. Economics of Data
    5.2. Data Dignity
6. Strategic adaptation
    6.1. Performative prediction
    6.2. Strategic classification
7. Data Valuation Researchers

What is your data worth?

Shapley Value & Cooperative Game Theory

Towards Efficient Data Valuation Based on the Shapley Value Ruoxi Jia & David Dao, Boxin Wang, Frances Ann Hubis, Nick Hynes, Nezihe Merve Gürel, Bo Li, Ce Zhang, Dawn Song, Costas J. Spanos 2019
Summary Jia et al. (2019) contribute theoretical and practical results for efficient methods for approximating the Shapley value (SV). They show that methods with a sublinear amount of model evaluations are possible and further reductions can be made for sparse SVs. Lastly, they introduce two practical SV estimation methods for ML tasks, one for uniformly stable learning algorithms and one for smooth loss functions.
Bibtex
@inproceedings{jia2019towards,
title={Towards efficient data valuation based on the shapley value},
author={Jia, Ruoxi and Dao, David and Wang, Boxin and Hubis, Frances Ann and Hynes, Nick and G{"u}rel, Nezihe Merve and Li, Bo and Zhang, Ce and Song, Dawn and Spanos, Costas J},
booktitle={The 22nd International Conference on Artificial Intelligence and Statistics},
pages={1167--1176},
year={2019},
organization={PMLR}
}
💻
Data Shapley: Equitable Valuation of Data for Machine Learning Amirata Ghorbani, James Zou 2019
Summary Ghorbani & Zou (2019) introduce (data) Shapley value to equitably measure the value of each training point to a supervised learners performance. They further outline several benefits of the Shapley value, e.g. being able to capture outliers or inform what new data to acquire, as well as develop Monte Carlo and gradient-based methods for its efficient estimation.
Bibtex
@inproceedings{ghorbani2019data,
title={Data shapley: Equitable valuation of data for machine learning},
author={Ghorbani, Amirata and Zou, James},
booktitle={International Conference on Machine Learning},
pages={2242--2251},
year={2019},
organization={PMLR}
}
💻
A Distributional Framework for Data Valuation Amirata Ghorbani, Michael P. Kim, James Zou 2020
Summary Ghorbani et al. (2020) formulate the Shapley value as a distributional quantity in the context of an underlying data distribution instead of a fixed dataset. They further introduce a novel sampling-based algorithm for the distributional Shapley value with strong approximation guarantees.
Bibtex
@inproceedings{ghorbani2020distributional,
title={A Distributional Framework for Data Valuation},
author={Ghorbani, Amirata, P. Kim, Michael and Zou, James},
booktitle={International Conference on Machine Learning},
year={2020}
}
💻
Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainability Christopher Frye, Colin Rowat, Ilya Feige 2020
Summary Frye et al. (2020) incorporate causality into the Shapley value framework. Importantly, their framework can handle any amount of causal knowledge and does not require the complete causal graph underlying the data.
Bibtex
@article{frye2020asymmetric,
title={Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainability},
author={Frye, Christopher and Rowat, Colin and Feige, Ilya},
journal={Advances in Neural Information Processing Systems},
volume={33},
year={2020}
}
🎥
Collaborative Machine Learning with Incentive-Aware Model Rewards Rachael Hwee Ling Sim, Yehong Zhang, Mun Choon Chan, Bryan Kian Hsiang Low 2020
Summary Sim et al. (2020) introduce a data valuation method with separate ML models as rewards based on the Shapley value and information gain on model parameters given its data. They further define several conditions for incentives such as Shapley fairness, stability, individual rationality, and group welfare, that are suitable for the freely replicable nature of their model reward scheme.
Bibtex
@inproceedings{sim2020collaborative,
title={Collaborative machine learning with incentive-aware model rewards},
author={Sim, Rachael Hwee Ling and Zhang, Yehong and Chan, Mun Choon and Low, Bryan Kian Hsiang},
booktitle={International Conference on Machine Learning},
pages={8927--8936},
year={2020},
organization={PMLR}
}
Validation free and replication robust volume-based data valuation Xinyi Xu, Zhaoxuan Wu, Chuan Sheng Foo, Bryan Kian Hsiang Low 2021
Summary Xu et al. (2021) propose using data diversity via robust volume for measuring the value of data. This removes the need for a validation set and allows for guarantees on replication robustness but suffers from the curse of dimensionality and may ignore useful information in the validation set.
Bibtex
@article{xu2021validation,
title={Validation free and replication robust volume-based data valuation},
author={Xu, Xinyi and Wu, Zhaoxuan and Foo, Chuan Sheng and Low, Bryan Kian Hsiang},
journal={Advances in Neural Information Processing Systems},
volume={34},
year={2021}
}
💻
Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning Yongchan Kwon, James Zou 2021
Summary Kwon & Zou (2022) introduce Beta Shapley, a generalization of Data Shapley by relaxing the efficiency axiom.
Bibtex
@article{kwon2021beta,
title={Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning},
author={Kwon, Yongchan and Zou, James},
journal={arXiv preprint arXiv:2110.14049},
year={2021}
}
Gradient-Driven Rewards to Guarantee Fairness in Collaborative Machine Learning Xinyi Xu, Lingjuan Lyu, Xingjun Ma, Chenglin Miao, Chuan Sheng Foo, Bryan Kian Hsiang Low 2021
Summary Xu et al. (2021) propose cosine gradient Shapley value to fairly evaluate the expected contribution of each agent's update in the federated learning setting removing the need for an auxiliary validation dataset. They further introduce a novel training-time gradient reward mechanism with a fairness guarantee.
Bibtex
@article{xu2021gradient,
title={Gradient driven rewards to guarantee fairness in collaborative machine learning},
author={Xu, Xinyi and Lyu, Lingjuan and Ma, Xingjun and Miao, Chenglin and Foo, Chuan Sheng and Low, Bryan Kian Hsiang},
journal={Advances in Neural Information Processing Systems},
volume={34},
pages={16104--16117},
year={2021}
}
Improving Cooperative Game Theory-based Data Valuation via Data Utility Learning Tianhao Wang, Yu Yang, Ruoxi Jia 2022
Summary Wang et al. (2022) propose a general framework to improve effectiveness of sampling-based Shapley value (SV) or Least core (LC) estimation heuristics. They propose learning to predict the performance of a learning algorithm (denoted data utility learning) and using this predictor to estimate learning performance without retraining for cheaper SV and LC estimation.
Bibtex
@article{wang2021improving,
title={Improving cooperative game theory-based data valuation via data utility learning},
author={Wang, Tianhao and Yang, Yu and Jia, Ruoxi},
journal={arXiv preprint arXiv:2107.06336},
year={2021}
}
Data Banzhaf: A Robust Data Valuation Framework for Machine Learning Jiachen T. Wang, Ruoxi Jia 2023
Summary Wang et al. (2023) propose using the Banzhaf value for data valuation, providing better robustness against noisy performance scores and an efficient estimate using Maximum Sample Reuse (MSR) principle
Bibtex
@InProceedings{pmlr-v206-wang23e, title={Data Banzhaf: A Robust Data Valuation Framework for Machine Learning},
author={Wang, Jiachen T. and Jia, Ruoxi},
booktitle={Proceedings of The 26th International Conference on Artificial Intelligence and Statistics},
pages={6388--6421},
year={2023},
editor={Ruiz, Francisco and Dy, Jennifer and van de Meent, Jan-Willem},
volume={206},
series={Proceedings of Machine Learning Research},
month={25--27 Apr},
publisher={PMLR},
pdf={https://proceedings.mlr.press/v206/wang23e/wang23e.pdf},
url={https://proceedings.mlr.press/v206/wang23e.html}
}
💻
A Multilinear Sampling Algorithm to Estimate Shapley Values Ramin Okhrati, Aldo Lipani 2021
Summary Okhrati and Lipani (2021) propose a new sampling method for Shapley values based on a multilinear extension technique as applied in game theory. It provides more accurate estimations of the Shapley values by reducing the variance of the sampling statistics.
Bibtex
@INPROCEEDINGS{9412511,
title={A Multilinear Sampling Algorithm to Estimate Shapley Values},
author={Okhrati, Ramin and Lipani, Aldo},
booktitle={2020 25th International Conference on Pattern Recognition (ICPR)},
year={2021}
}
💻
If You Like Shapley Then You’ll Love the Core Yan, T., and Procaccia, A. D. 2021
Summary Yan and Procaccia (2021) propose an alternative method for credit assignment in data valuation. They use the least core, which can be computed efficiently.
Bibtex
@article{Yan_Procaccia_2021,
title={If You Like Shapley Then You’ll Love the Core},
author={Yan, Tom and Procaccia, Ariel D.},
journal={Proceedings of the AAAI Conference on Artificial Intelligence},
year={2021}
}
CS-Shapley: Class-wise Shapley Values for Data Valuation in Classification Schoch, Stephanie, Haifeng Xu, and Yangfeng Ji 2022
Summary Schoch et al. (2022) propose a new Shapley value that discriminates between training instances' in-class and out-of-class contributions.
Bibtex
@inproceedings{schoch2022csshapley,
title={{CS}-Shapley: Class-wise Shapley Values for Data Valuation in Classification},
author={Stephanie Schoch and Haifeng Xu and Yangfeng Ji},
booktitle={Advances in Neural Information Processing Systems},
year={2022}
}
💻

Efficient algorithms

Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms Ruoxi Jia, David Dao, Boxin Wang, Frances Ann Hubis, Nezihe Merve Gurel, Bo Li, Ce Zhang, Costas J. Spanos, Dawn Song 2019
Summary Jia et al. (2019) present algorithms to compute the Shapley value exactly in quasi-linear time and approximations in sublinear time for k-nearest-neighbor models. They empirically evaluate their algorithms at scale and extend them to several other settings.
Bibtex
@article{jia12efficient,
title={Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms},
author={Jia, Ruoxi and Dao, David and Wang, Boxin and Hubis, Frances Ann and Gurel, Nezihe Merve and Zhang, Bo Li4 Ce and Song, Costas Spanos1 Dawn},
journal={Proceedings of the VLDB Endowment},
volume={12},
number={11}
}
💻
Efficient computation and analysis of distributional Shapley values Yongchan Kwon, Manuel A. Rivas, James Zou 2021
Summary Kwon et al. (2021) develop tractable analytic expressions for the distributional data Shapley value for linear regression, binary classification, and non-parametric density estimation as well as new efficient methods for its estimation.
Bibtex
@inproceedings{kwon2021efficient,
title={Efficient computation and analysis of distributional Shapley values},
author={Kwon, Yongchan and Rivas, Manuel A and Zou, James},
booktitle={International Conference on Artificial Intelligence and Statistics},
pages={793--801},
year={2021},
organization={PMLR}
}
💻

Benchmarks, Criticism & Relaxations

Scalability vs. Utility: Do We Have to Sacrifice One for the Other in Data Importance Quantification? Ruoxi Jia, Fan Wu, Xuehui Sun, Jiacen Xu, David Dao, Bhavya Kailkhura, Ce Zhang, Bo Li, Dawn Song 2021
Summary Jia et al. (2021) perform a theoretical analysis on the differences between leave-one-out-based and Shapley value-based methods as well as an empirical study across several ML tasks investigating the two aforementioned methods as well as exact Shapley value-based methods and Shapley over KNN Surrogates.
Bibtex
@misc{jia2021scalability,
title={Scalability vs. Utility: Do We Have to Sacrifice One for the Other in Data Importance Quantification?},
author={Ruoxi Jia and Fan Wu and Xuehui Sun and Jiacen Xu and David Dao and Bhavya Kailkhura and Ce Zhang and Bo Li and Dawn Song},
year={2021},
eprint={1911.07128},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
💻
Shapley values for feature selection: The good, the bad, and the axioms Daniel Fryer, Inga Strümke, Hien Nguyen 2021
Summary Fryer et al. (2021) calls into question the appropriateness of using the Shapley value for feature selection and advise caution against the magical thinking that presenting its abstract general axioms as "favourable and fair" may introduce. They further point out that the four axioms of "efficiency", "null player", "symmetry", and "additivity" do not guarantee that the Shapley value is suited to feature selection and may sometimes even imply the opposite.
Bibtex
@misc{fryer2021shapley,
title={Shapley values for feature selection: The good, the bad, and the axioms},
author={Daniel Fryer and Inga Strümke and Hien Nguyen},
year={2021},
eprint={2102.10936},
archivePrefix={arXiv},
primaryClass={cs.LG}
}

Influence functions & LOO

Understanding Black-box Predictions via Influence Functions Pang Wei Koh, Percy Liang 2017
Summary Koh & Liang (2017) introduce the use of influence functions, a technique borrowed from robust statistics, to identify training points most responsible for a model's given prediction without needing to retrain. They further develop a simple and efficient implementation of influence functions that scales to large ML settings.
Bibtex
@inproceedings{koh2017understanding,
title={Understanding black-box predictions via influence functions},
author={Koh, Pang Wei and Liang, Percy},
booktitle={International Conference on Machine Learning},
pages={1885--1894},
year={2017},
organization={PMLR}
}
💻 🎥
On the accuracy of influence functions for measuring group effects Pang Wei Koh*, Kai-Siang Ang*, Hubert H. K. Teo*, and Percy Liang 2019
Summary Koh et al. (2019) study influence functions to measure effects of large groups of training points instead of individual points. They empirically find a correlation and often underestimation between predicted and actual effects and theoretically show that this need not hold in general, realistic settings.
Bibtex
@article{koh2019accuracy,
title={On the accuracy of influence functions for measuring group effects},
author={Koh, Pang Wei and Ang, Kai-Siang and Teo, Hubert HK and Liang, Percy},
journal={arXiv preprint arXiv:1905.13289},
year={2019}
}
💻 🎥
Scaling Up Influence Functions Schioppa, Andrea, Polina Zablotskaia, David Vilar, and Artem Sokolov 2022
Summary Schioppa et al. (2022) propose a new method to scale the computation of influence functions for large neural networks using the Arnoldi iteration. With this, they achieve successful implementation of influence functions on full-size Transformer models with hundreds of millions of parameters.
Bibtex
@inproceedings{schioppa2022scaling,
title={Scaling Up Influence Functions},
author={Schioppa, Andrea and Zablotskaia, Polina and Vilar, David and Sokolov, Artem},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
year={2022}
}
💻
Studying large language model generalization with influence functions Grosse, Roger and Bae, Juhan and Anil, Cem and Elhage, Nelson and Tamkin, Alex and Tajdini, Amirhossein and Steiner, Benoit and Li, Dustin and Durmus, Esin and Perez, Ethan and others 2023
Summary Grosse et al. (2023) use a method known as EK-FAC to approximate the Hessian of the loss of large language models. They apply this technique to study influence functions on large language models, up to 50 billion parameters.
Bibtex
@article{grosse2023studying,
title={Studying large language model generalization with influence functions},
author={Grosse, Roger and Bae, Juhan and Anil, Cem and Elhage, Nelson and Tamkin, Alex and Tajdini, Amirhossein and Steiner, Benoit and Li, Dustin and Durmus, Esin and Perez, Ethan and others},
journal={arXiv preprint arXiv:2308.03296},
year={2023}
}

Reinforcement Learning

Data Valuation using Reinforcement Learning Jinsung Yoon, Sercan Ö Arık, Tomas Pfister 2020
Summary Yoon et al. (2020) propose using reinforcement learning for data valuation to learn data values jointly with the predictor model.
Bibtex
@inproceedings{49189,
title={Data Valuation using Reinforcement Learning},
author={Jinsung Yoon and Sercan Arik and Tomas Pfister},
year={2020}
}
💻 🎥

Deep Neural Networks

DAVINZ: Data Valuation using Deep Neural Networks at Initialization Zhaoxuan Wu, Yao Shu, Bryan Kian Hsiang Low 2022
Summary Wu et al. (2022) introduce a validation-based and training-free method for efficient data valuation with large and complex deep neural networks (DNNs). They derive and exploit a domain-aware generalization bound for DNNs to characterize their performance without training and uses this bound as the scoring function while keeping conventional techniques such as Shapley values as the valuation function.
Bibtex
@inproceedings{wu2022davinz,
title={DAVINZ: Data Valuation using Deep Neural Networks at Initialization},
author={Wu, Zhaoxuan and Shu, Yao and Low, Bryan Kian Hsiang},
booktitle={International Conference on Machine Learning},
pages={24150--24176},
year={2022},
organization={PMLR}
}
🎥

Out-of-Bag score

Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Value Yongchan Kwon, James Zou 2023
Summary Kwon et al. (2023) propose using the out-of-bag estimate of a bagging estimator for computationally efficient data valuation.
Bibtex
@inproceedings{DBLP:conf/icml/Kwon023, 
author={Yongchan Kwon and James Zou},
editor={Andreas Krause and Emma Brunskill and Kyunghyun Cho and Barbara Engelhardt and Sivan Sabato and Jonathan Scarlett},
title={Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Value},
booktitle={International Conference on Machine Learning, {ICML} 2023, 23-29 July 2023, Honolulu, Hawaii, {USA}},
series={Proceedings of Machine Learning Research},
volume={202},
pages={18135--18152},
publisher={{PMLR}},
year={2023},
url={https://proceedings.mlr.press/v202/kwon23e.html},
timestamp={Mon, 28 Aug 2023 17:23:08 +0200},
biburl={https://dblp.org/rec/conf/icml/Kwon023.bib},
bibsource={dblp computer science bibliography, https://dblp.org}
}
💻 🎥

Task Agnostic

Fundamentals of Task-Agnostic Data Valuation Mohammad Mohammadi Amiri, Frederic Berdoz, Ramesh Raskar 2023
SummaryThis paper addresses the challenge of valuing data without specific task assumptions, focusing on task-agnostic data valuation. It discusses valuing a data seller's dataset from a buyer's perspective without validation requirements. The approach involves estimating statistical differences through diversity and relevance measures without needing the raw data, and designing queries that maintain the seller's blindness to the buyer's raw data. The work is significant for practical scenarios where utility metrics like test accuracy on a validation set are not feasible.
Bibtex
@article{Amiri2023FundamentalsOT,
title={Fundamentals of Task-Agnostic Data Valuation},
author={Mohammad Mohammadi Amiri and Frederic Berdoz and Ramesh Raskar},
journal={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={37},
pages={9226-9234},
year={2023},
doi={10.1609/aaai.v37i8.26106}
}

Benchmarks

OpenDataVal: a Unified Benchmark for Data Valuations Kevin Jiang, Weixin Liang, James Zou, Yongchan Kwon 2023
Summary Jiang et al. (2023) provides a Python library to build and test data evaluators across different datasets, data evaluators, models, and new benchmarks.
Bibtex
@article{jiang2023opendataval,
title={OpenDataVal: a Unified Benchmark for Data Valuation},
author={Kevin Fu Jiang and Weixin Liang and James Zou and Yongchan Kwon},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
year={2023},
url={https://openreview.net/forum?id=eEK99egXeB}
}
💻 🎥

Libraries

influenciae Deel-AI 2023
Summary A stable implementation of influence functions in tensorflow.
Bibtex
💻
pyDVL appliedAI Institute 2023
Summary A library of stable and efficient implementations of algorithms for computing Shapley values and influence functions in pytorch.
Bibtex
💻

Surveys

Data Valuation in Machine Learning: “Ingredients”, Strategies, and Open Challenges Rachael Hwee Ling Sim*, Xinyi Xu*, Bryan Kian Hsiang Low 2022
Summary Sim et al. (2022) present a technical survey of data valuation and its "ingredients" and properties. The paper outlines common desiderata as well as some open research challenges.
Bibtex
@inproceedings{sim2022data,
title={Data valuation in machine learning:“ingredients”, strategies, and open challenges},
author={Sim, Rachael Hwee Ling and Xu, Xinyi and Low, Bryan Kian Hsiang},
booktitle={Proc. IJCAI},
year={2022}
}
🎥

Designing data marketplaces

Data market system designs

A demonstration of sterling: a privacy-preserving data marketplace Nick Hynes, David Dao, David Yan, Raymond Cheng, Dawn Song 2018
Bibtex
@article{hynes2018demonstration,
title={A Demonstration of Sterling: A Privacy-Preserving Data Marketplace},
author={Hynes, Nick and Dao, David and Yan, David and Cheng, Raymond and Song, Dawn},
journal={Proceedings of the VLDB Endowment},
volume={11},
number={12},
year={2018}
}
DataBright: Towards a Global Exchange for Decentralized Data Ownership and Trusted Computation David Dao, Dan Alistarh, Claudiu Musat, Ce Zhang 2018
Bibtex
@article{dao2018databright,
title={Databright: Towards a global exchange for decentralized data ownership and trusted computation},
author={Dao, David and Alistarh, Dan and Musat, Claudiu and Zhang, Ce},
journal={arXiv preprint arXiv:1802.04780},
year={2018}
}
A Marketplace for Data: An Algorithmic Solution Anish Agarwal, Munther Dahleh, Tuhin Sarkar 2019
Bibtex
@inproceedings{agarwal2019marketplace,
title={A marketplace for data: An algorithmic solution},
author={Agarwal, Anish and Dahleh, Munther and Sarkar, Tuhin},
booktitle={Proceedings of the 2019 ACM Conference on Economics and Computation},
pages={701--726},
year={2019}
}
Computing a Data Dividend Eric Bax 2019
Bibtex
@misc{bax2019computing,
title={Computing a Data Dividend},
author={Eric Bax},
year={2019},
eprint={1905.01805},
archivePrefix={arXiv},
primaryClass={cs.GT}
}
Incentivizing Collaboration in Machine Learning via Synthetic Data Rewards Sebastian Shenghong Tay, Xinyi Xu, Chuan Sheng Foo, Bryan Kian Hsiang Low 2021
Bibtex
@article{tay2021incentivizing,
title={Incentivizing Collaboration in Machine Learning via Synthetic Data Rewards},
author={Tay, Sebastian Shenghong and Xu, Xinyi and Foo, Chuan Sheng and Low, Bryan Kian Hsiang},
journal={arXiv preprint arXiv:2112.09327},
year={2021}
}

Automatic data compliance

Data Capsule: A New Paradigm for Automatic Compliance with Data Privacy Regulations Lun Wang, Joseph P. Near, Neel Somani, Peng Gao, Andrew Low, David Dao, Dawn Song 2019
Bibtex
@misc{wang2019data,
title={Data Capsule: A New Paradigm for Automatic Compliance with Data Privacy Regulations},
author={Lun Wang and Joseph P. Near and Neel Somani and Peng Gao and Andrew Low and David Dao and Dawn Song},
year={2019},
eprint={1909.00077},
archivePrefix={arXiv},
primaryClass={cs.CY}
}
💻

Data valuation applications

A Principled Approach to Data Valuation for Federated Learning Tianhao Wang, Johannes Rausch, Ce Zhang, Ruoxi Jia, Dawn Song 2020
Bibtex
@misc{wang2020principled,
title={A Principled Approach to Data Valuation for Federated Learning},
author={Tianhao Wang and Johannes Rausch and Ce Zhang and Ruoxi Jia and Dawn Song},
year={2020},
eprint={2009.06192},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
Data valuation for medical imaging using Shapley value and application to a large-scale chest X-ray dataset Siyi Tang, Amirata Ghorbani, Rikiya Yamashita, Sameer Rehman, Jared A Dunnmon, James Zou, Daniel L Rubin 2021
Bibtex
@article{tang2021data,
title={Data valuation for medical imaging using Shapley value and application to a large-scale chest X-ray dataset},
author={Tang, Siyi and Ghorbani, Amirata and Yamashita, Rikiya and Rehman, Sameer and Dunnmon, Jared A and Zou, James and Rubin, Daniel L},
journal={Scientific reports},
volume={11},
number={1},
pages={1--9},
year={2021},
publisher={Nature Publishing Group}
}
Efficient and Fair Data Valuation for Horizontal Federated Learning Shuyue Wei, Yongxin Tong, Zimu Zhou, Tianshu Song 2020
SummaryAvailability of big data is crucial for modern machine learning applications and services. Federated learning is an emerging paradigm to unite different data owners for machine learning on massive data sets without worrying about data privacy. Yet data owners may still be reluctant to contribute unless their data sets are fairly valuated and paid. In this work, the authors adapt Shapley value, a widely used data valuation metric to valuating data providers in federated learning. Prior data valuation schemes for machine learning incur high computation cost because they require training of extra models on all data set combinations. For efficient data valuation, the authors approximately construct all the models necessary for data valuation using the gradients in training a single model, rather than train an exponential number of models from scratch. On this basis, they devise three methods for efficient contribution index estimation. Evaluations show that their methods accurately approximate the contribution index while notably accelerating its calculation.
Bibtex
@inbook{wei2020efficient,
title={Efficient and fair data valuation for horizontal federated learning},
author={Wei, Shuyue and Tong, Yongxin and Zhou, Zimu and Song, Tianshu},
year={2020},
booktitle={Federated Learning: Privacy and Incentive},
pages={139--152},
publisher={Springer}
}
Improving Fairness for Data Valuation in Horizontal Federated Learning Zhenan Fan, Huang Fang, Zirui Zhou, Jian Pei, Michael P. Friedlander, Changxin Liu, Yong Zhang 2020
SummaryFederated learning is an emerging decentralized machine learning scheme that allows multiple data owners to work collaboratively while ensuring data privacy. This paper focuses on fairness in data valuation within federated learning. The authors propose a new measure called completed federated Shapley value to improve the fairness of federated Shapley value. This approach leverages the concepts and tools from optimization and provides both theoretical analysis and empirical evaluation to verify the improvement in fairness.
Bibtex
@misc{fan2020improving,
title={Improving Fairness for Data Valuation in Horizontal Federated Learning},
author={Zhenan Fan and Huang Fang and Zirui Zhou and Jian Pei and Michael P. Friedlander and Changxin Liu and Yong Zhang},
year={2020},
eprint={2109.09046},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
Data Valuation for Vertical Federated Learning: An Information-Theoretic Approach Xiao Han, Leye Wang, Junjie Wu 2021
SummaryFederated learning (FL) is a machine learning paradigm that enables privacy-preserving cross-party data collaboration. This work introduces "FedValue," the first privacy-preserving, task-specific, model-free data valuation method for vertical FL tasks. It incorporates Shapley-CMI, an information-theoretic metric, for assessing data values from a game-theoretic perspective. The paper also proposes a novel server-aided federated computation mechanism and techniques to accelerate Shapley-CMI computation. Extensive experiments demonstrate the effectiveness and efficiency of FedValue.
Bibtex
@misc{han2021datavaluation,
title={Data Valuation for Vertical Federated Learning: An Information-Theoretic Approach},
author={Xiao Han and Leye Wang and Junjie Wu},
year={2021},
eprint={URL or DOI link TBD},
}
Towards More Efficient Data Valuation in Healthcare Federated Learning Using Ensembling Sourav Kumar, A. Lakshminarayanan, Ken Chang, Feri Guretno, Ivan Ho Mien, Jayashree Kalpathy-Cramer, Pavitra Krishnaswamy, Praveer Singh 2021
SummaryThis paper addresses the challenge of data valuation in federated learning within healthcare. The authors propose a method called SaFE (Shapley Value for Federated Learning using Ensembling), which is designed to be efficient in settings where the number of contributing institutions is manageable. SaFE approximates the Shapley value using gradients from training a single model and develops methods for efficient contribution index estimation. This approach is particularly relevant in medical imaging where data heterogeneity is common and fast, accurate data valuation is necessary for multi-institutional collaborations.
Bibtex
@article{Kumar2021TowardsME,
title={Towards More Efficient Data Valuation in Healthcare Federated Learning Using Ensembling},
author={Sourav Kumar and A. Lakshminarayanan and Ken Chang and Feri Guretno and Ivan Ho Mien and Jayashree Kalpathy-Cramer and Pavitra Krishnaswamy and Praveer Singh},
journal={ArXiv},
year={2021},
volume={abs/2209.05424}
}

Data markets and society

Economics of Data

Nonrivalry and the Economics of Data Charles I. Jones, Christopher Tonetti 2019
Bibtex
@article{10.1257/aer.20191330,
Author = {Jones, Charles I. and Tonetti, Christopher},
Title = {Nonrivalry and the Economics of Data},
Journal = {American Economic Review},
Volume = {110},
Number = {9},
Year = {2020},
Month = {September},
Pages = {2819-58},
DOI = {10.1257/aer.20191330},
URL = {https://www.aeaweb.org/articles?id=10.1257/aer.20191330}
}

Data Dignity

Chapter 5: Data as Labor, Radical Markets Eric A. Posner and E Glen Weyl 2019
Bibtex
@book{posner2019radical,
title={Radical Markets},
author={Posner, Eric A and Weyl, E Glen},
year={2019},
publisher={Princeton University Press}
}
Should We Treat Data as Labor? Moving beyond "Free" Imanol Arrieta-Ibarra, Leonard Goff, Diego Jiménez-Hernández, Jaron Lanier, E. Glen Weyl 2018
Bibtex
@article{10.1257/pandp.20181003,
Author = {Arrieta-Ibarra, Imanol and Goff, Leonard and Jiménez-Hernández, Diego and Lanier, Jaron and Weyl, E. Glen},
Title = {Should We Treat Data as Labor? Moving beyond "Free"},
Journal = {AEA Papers and Proceedings},
Volume = {108},
Year = {2018},
Month = {May},
Pages = {38-42},
DOI = {10.1257/pandp.20181003},
URL = {https://www.aeaweb.org/articles?id=10.1257/pandp.20181003}
}

Strategic adaptation

Performative prediction

Performative Prediction Juan Perdomo, Tijana Zrnic, Celestine Mendler-Dünner, Moritz Hardt 2020
Summary Perdomo et al. (2020) introduce the concept of "performative prediction" dealing with predictions that influence the target they aim to predict, e.g. through taking actions based on the predictions, causing a distribution shift. The authors develop a risk minimization framework for performative prediction and introduce the equilibrium notion of performative stability where predictions are calibrated against future outcomes that manifest from acting on the prediction.
Bibtex
@inproceedings{perdomo2020performative,
title={Performative prediction},
author={Perdomo, Juan and Zrnic, Tijana and Mendler-D{"u}nner, Celestine and Hardt, Moritz},
booktitle={International Conference on Machine Learning},
pages={7599--7609},
year={2020},
organization={PMLR}
}
Stochastic Optimization for Performative Prediction Celestine Mendler-Dünner, Juan Perdomo, Tijana Zrnic, Moritz Hardt 2020
Summary Mendler-Dünner et al. (2020) look at stochastic optimization for performative prediction and prove convergence rates for greedily deploying models after each stochastic update (which may cause distribution shift affecting convergence to a stability point) or lazily deploying the model after several updates.
Bibtex
@article{mendler2020stochastic,
title={Stochastic optimization for performative prediction},
author={Mendler-D{"u}nner, Celestine and Perdomo, Juan and Zrnic, Tijana and Hardt, Moritz},
journal={Advances in Neural Information Processing Systems},
volume={33},
pages={4929--4939},
year={2020}
}

Strategic classification

Strategic Classification is Causal Modeling in Disguise John Miller, Smitha Milli, Moritz Hardt 2020
Summary Miller et al. (2020) argue that strategic classication involves causal modelling and designing incentives for improvement requires solving a non-trivial causal inference problem. The authors provide a distinction between gaming and improvement as well as provide a causal framework for strategic adaptation.
Bibtex
@inproceedings{miller2020strategic,
title={Strategic classification is causal modeling in disguise},
author={Miller, John and Milli, Smitha and Hardt, Moritz},
booktitle={International Conference on Machine Learning},
pages={6917--6926},
year={2020},
organization={PMLR}
}
Alternative Microfoundations for Strategic Classification Meena Jagadeesan, Celestine Mendler-Dünner, Moritz Hardt 2021
Summary Jagadeesan et al. (2021) show that standard microfoundations in strategic classification, that typically uses individual-level behaviour to deduce aggregate-level responses, can lead to degenerate behaviour in aggregate: discontinuities in the aggregate response, stable points ceasing to exist, and maximizing social burden. The authors introduce a noisy response model inspired by performative prediction that mitigates these limitations for binary classification.
Bibtex
@inproceedings{jagadeesan2021alternative,
title={Alternative microfoundations for strategic classification},
author={Jagadeesan, Meena and Mendler-D{"u}nner, Celestine and Hardt, Moritz},
booktitle={International Conference on Machine Learning},
pages={4687--4697},
year={2021},
organization={PMLR}
}

Data Valuation Researchers

Name Institute h-index
Costas Spanos University of California, Berkeley 61
Jinsung Yoon Google Cloud AI 33
Tomas Pfister Google Cloud AI 39
Amirata Ghorbani Stanford 18
James Zou Stanford 64
Nektaria Tryfona Virginia Tech 27
Rachael Hwee Ling Sim National University of Singapore 4
Bryan Kian Hsiang Low National University of Singapore 38
Dawn Song University of California, Berkeley 142
Zhaoxuan Wu National University of Singapore 4
Xinyi Xu National University of Singapore 8
Tianhao Wang University of Virginia 18
José González Cabañas UC3M-Santander Big Data Institute 7
Ruben Cuevas Rumin Universidad Carlos III de Madrid 26
Jiachen T. Wang Princeton University 9
Bohong Wang Tsinghua University 6
Yongchan Kwon Columbia University 10
Siyi Tang Artera 8
Li Xiong Emory University 52
Jessica Vitak University of Maryland 49
Katie Chamberlain Kritikos University of Illinois at Urbana-Champaign 6
Zhenan Fan Huawei Technologies Canada 6
Shuyue Wei Beihang University 4
Hannah Stein Saarland University 3
Wolfgang Maass Saarland University 26
Mohammad Mohammadi Amiri Rensselaer Polytechnic Institute 18
Ramesh Raskar MIT 103
Konstantin D. Pandl Karlsruhe Institute of Technolgoy 6
Ali Sunyaev Karlsruhe Institute of Technolgoy 43
Ludovico Boratto University of Cagliari 25
han xiao 70
Junjie Wu Center for High Pressure Science & Technology Advanced Research 55
Xiao Tian National University of Singapore 1
Kean Birch Institute for Technoscience & Society 40
Callum Ward Uppsala University 10
Praveer Singh University of Colorado School of Medicine 19
Anran Xu Shanghai Jiao Tong University 2
Guihai Chen 67
Andre Esteva Co-Founder & CEO, Artera 23
Prateek Mittal Princeton University 55
Hyeontaek Oh Institute for IT Convergence 9
Lingjiao Chen Stanford 13
Xiangyu Chang Xi'an Jiaotong University 17
Hoang Anh Just Virginia Tech 3
David Dao ETH 13
Mark Mazumder Harvard 12
Vijay Janapa Reddi Harvard 46
Sabri Eyuboglu Stanford 6
Wenqian Li National University of Singapore 2