Data Shapley: Equitable Valuation of Data for Machine Learning

By Amirata Ghorbani, James Y. Zou. 2019

1,148 citations recorded by the source corpus.

As data becomes the fuel driving technological and economic growth, a fundamental challenge is how to quantify the value of data in algorithmic predictions and decisions. For example, in healthcare and consumer markets, it has been suggested that individuals should be compensated for the data that they generate, but it is not clear what is an equitable valuation for individual data. In this work, we develop a principled framework to address data valuation in the context of supervised machine learning. Given a learning algorithm trained on $n$ data points to produce a predictor, we propose data Shapley as a metric to quantify the value of each training datum to the predictor performance. Data Shapley value uniquely satisfies several natural properties of equitable data valuation. We develop Monte Carlo and gradient-based methods to efficiently estimate data Shapley values in practical settings where complex learning algorithms, including neural networks, are trained on large datasets. In addition to being equitable, extensive experiments across biomedical, image and synthetic data demonstrate that data Shapley has several other benefits: 1) it is more powerful than the popular leave-one-out or leverage score in providing insight on what data is more valuable for a given learning task; 2) low Shapley value data effectively capture outliers and corruptions; 3) high Shapley value data inform what type of new data to acquire to improve the predictor.

Subfields: Artificial Intelligence, Management Science and Operations Research, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Management Science and Operations Research

In Evaluation & Benchmarks

Citations in the corpus

Understanding Dataset Difficulty with V-Usable Information (2021, neutral): “Influence functions (Koh & Liang, 2017), forgetting events (Toneva et al., 2018), and the Data Shapley (Ghorbani & Zou, 2019; Jia et al., 2019) can all be used to assign point-wise estimates of importance to data instances based on their contribution to the decision boundary.”
Foundation Models and Fair Use (2023, neutral): “…attribution refers to methods that assign attribution scores to training examples to understand the contribution of individual examples (or group of examples) to (test-time) model predictions (Koh & Liang, 2017; Ghorbani & Zou, 2019; Jia et al., 2019; Pezeshkpour et al., 2021; Ilyas et al., 2022).”
Online Class-Incremental Continual Learning with Adversarial Shapley Value (2020, methodology): “Fortunately, the existing research literature already provides both a precise and efficient answer to this question viewed through the lens of Shapley data valuation for KNN classifiers (Jia et al. 2019a; Ghorbani and Zou 2019; Jia et al. 2019b).”
DeltaGrad: Rapid retraining of machine learning models (2020, methodology): “Our method can be also used to evaluate the importance of training samples (see Cook (1977) and the follow-up works such as Ghorbani & Zou (2019)).”
Making AI Forget You: Data Deletion in Machine Learning (2019, methodology): “To name a few examples, it could be used to speed-up leave-oneout-cross-validation [2], support a user data marketplace [75, 80], or identify important or valuable datapoints within a model [37].”
Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics (2020, methodology): “Inﬂuence functions (Koh and Liang, 2017), forgetting events (Toneva et al., 2018), cross validation (Chen et al., 2019), Shapely values (Ghorbani and Zou, 2019), and the area-under-margin metric (Pleiss et al., 2020) have all been used to identify mislabeled examples.”
Algorithms to estimate Shapley value feature attributions (2022, background): “Other examples include games where the value is the maximum flow of attention weights in transformer models [80], where the players are analogous to samples in the training data [81], where the players are analogous to neurons in a deep model [82], and where the players are analogous to edges in a c”
DataComp: In search of the next generation of multimodal datasets (2023, methodology): “Such methods include influence functions, a classic technique from robust statistics [52, 31] that uses a second-order Taylor expansion to approximate the effect of removing a training point on the learned model parameters [74, 75, 53, 48], as well as methods that fit attribution functions directly ”
Overview and Importance of Data Quality for Machine Learning Tasks (2020, neutral): “Data Valuation: In this topic, we will cover following research papers [7, 18] to communicate how the field of valuation of data points evolved over the period and how it can be calculated using different ways e.”
Adversarial Shapley Value Experience Replay for Task-Free Continual Learning (2020, methodology): “ASER is inspired by the Shapley value (SV) [30] used in cooperative game theory to fairly distribute total gains to all players — in our CL setting, we use the SV to determine the contribution of memory samples to learning performance [11, 12, 13].”
Estimating Training Data Influence by Tracking Gradient Descent (2020, background): “(Jia et al., 2019; Ghorbani & Zou, 2019) identify the inﬂuence of training examples on the overall accuracy of the model.”
On the Accuracy of Influence Functions for Measuring Group Effects (2019, methodology): “They can also be used to approximate Shapley values, which are a different but related way of measuring the effect of data points; see, e.g., Jia et al. (2019) and Ghorbani and Zou (2019).”
Secure, privacy-preserving and federated machine learning in medical imaging (2020, background): “…over one’s identity 100 and new techniques to quantify and track the added value of individual datasets with respect to algorithm performance will strengthen the notion of private data as a scarce and valuable resource within an evolving data economy 101 currently experiencing oversupply 102 .”
The future of digital health with federated learning (2020, methodology): “The measurements could then be used to determine relevant compensation and establish a revenue model among the participants [61].”
TRAK: Attributing Model Behavior at Scale (2023, positive): “On the other hand, sampling-based methods such as empirical influence functions (Feldman & Zhang, 2020), Shapley value estimators (Ghorbani & Zou, 2019) or datamodels (Ilyas et al., 2022) are more successful at accurately attributing predictions to training data but require training thousands (or…”
Advances, challenges and opportunities in creating data for trustworthy AI (2022, methodology): “This can be done effectively using recent methods such as data Shapley scores or influence approximations 48–51 .”
The Shapley Value in Machine Learning (2022, methodology): “, 2019] Validation loss Restricted Monte Carlo sampling O( √ |N | log |N |(2)) [Ghorbani et al., 2019] Validation loss Monte Carlo sampling O(|N |) [Shim et al.”