liveBook · Manning

Coalition

As explained in Ensemble Methods for Machine Learning

A coalition is a subset of cooperating players whose collective value is evaluated in a cooperative game.

A coalition is any subset of cooperating players whose collective effectiveness or value is evaluated. In SHAP, coalitions reveal how a feature or player changes a prediction or total reward when included. section 9.4.2

Prize-sharing example

Four data scientists form coalitions while pursuing a $20,000 Kaggle prize. Ava alone is worth $7,000; Ava, Ben, and Dev together are worth $13,000; all four form the grand coalition worth $20,000. section 9.4.2

A coalition table organizes the evaluation:

  • List every subset, including the empty coalition and the grand coalition. section 9.4.2
  • Assign each coalition a characteristic-function value representing its effectiveness or worth. section 9.4.2
  • Compare corresponding coalitions with and without one player to measure that player’s contribution. section 9.4.2
Figure 9.16 All possible coalitions of Ava, Ben, Cam, and Dev, and their corresponding values (in units of $1,000). The last coalition contains all four friends and has a value of $20,000, the total prize money. There is one coalition of size 0, four coalitions of size 1, six coalitions of size 2, four coalitions of size 3, and one coalition of size 4. This table is called the characteristic function.
CH09_F16_Kunapuli

The figure shows the complete characteristic function: one empty coalition, four single-player coalitions, six pairs, four triples, and the four-player grand coalition. section 9.4.2

How Ensemble Methods for Machine Learning covers it

How deep it goes
Defines coalition enumeration and valuation, then connects with-and-without comparisons to Shapley-value attribution.
Examples
Uses four data scientists, a Kaggle prize, coalition-value tables, and a four-player grand coalition.
What it assumes
Assumes familiarity with cooperating players and collective value, while defining coalition and grand coalition explicitly.
What you can do after
Evaluate a player’s contribution by comparing coalition values with and without that player.
How it connects
Connects coalitions to cooperative games, characteristic functions, marginal contributions, and Shapley values.

This explainer was generated by Manning's knowledge graph from the text of Ensemble Methods for Machine Learning. The section links point to the parts of the book it draws on.