Abstract:Shapley value is a concept in cooperative game theory for measuring the contribution of each participant, which was named in honor of Lloyd Shapley. Shapley value has been recently applied in data marketplaces for compensation allocation based on their contribution to the models. Shapley value is the only value division scheme used for compensation allocation that meets three desirable criteria: group rationality, fairness, and additivity. In cooperative game theory, the marginal contribution of each contributor to each coalition is a nonnegative value. However, in machine learning model training, the marginal contribution of each contributor (data tuple) to each coalition (a set of data tuples) can be a negative value, i.e., the accuracy of the model trained by a dataset with an additional data tuple can be lower than the accuracy of the model trained by the dataset only. In this paper, we investigate the problem of how to handle the negative marginal contribution when computing Shapley value. We explore three philosophies: 1) taking the original value (Original Shapley Value); 2) taking the larger of the original value and zero (Zero Shapley Value); and 3) taking the absolute value of the original value (Absolute Shapley Value). Experiments on Iris dataset demonstrate that the definition of Absolute Shapley Value significantly outperforms the other two definitions in terms of evaluating data importance (the contribution of each data tuple to the trained model).

A Value for Weighted Two-level Coalition Structure

The Equal Surplus Division Value for Cooperative Games with a Level Structure

Absolute Shapley Value

Cooperative networks and f-Shapley value

A note on the per capita Shapley support levels value

Fuzzy Harsanyi solutions for fuzzy level structure games with multi weight systems

Equal surplus sharing values for games with coalition structures

Coalitional Manipulations and Immunity of the Shapley Value

Improved Shapley Values Based on Players’ Least Square Contributions and Their Applications in the Collaborative Profit Sharing of the Rural E-commerce

On egalitarian values for cooperative games with level structures

The Shapley Value of Cooperative Game with Stochastic Payoffs

Shapley Meets Shapley

Formation of coalition structures as a non-cooperative game 2: applications

Shapley Value Approximation Based on Complementary Contribution

Counterfactual Explanation of Shapley Value in Data Coalitions

On egalitarian values for cooperative games with a priori unions

A value for cooperative games on simplicial complexes with a filtration

Necessary players and values

Formation of coalition structures as a non-cooperative game

Two families of values for global cooperative games

An algorithm of coalition structure generation with given required bound based on cardinality structure