create_coalition_table: Define coalitions, and fetch additional information about...

In NorskRegnesentral/shapr: Prediction Explanation with Dependence-Aware Shapley Values

Define coalitions, and fetch additional information about each unique coalition

Description

Define coalitions, and fetch additional information about each unique coalition

Usage

create_coalition_table(
  m,
  exact = TRUE,
  n_coalitions = 200,
  weight_zero_m = 10^6,
  paired_shap_sampling = TRUE,
  prev_coal_samples = NULL,
  prev_coal_samples_n_unique = NULL,
  n_samps_scale = 10,
  coal_feature_list = as.list(seq_len(m)),
  approach0 = "gaussian",
  kernelSHAP_reweighting = "none",
  dt_valid_causal_coalitions = NULL
)

Arguments

m	Positive integer. Total number of features/groups.
exact	Logical. If TRUE all 2^m coalitions are generated, otherwise a subsample of the coalitions is used.
n_coalitions	Positive integer. Note that if exact = TRUE, n_coalitions is ignored.
weight_zero_m	Numeric. The value to use as a replacement for infinite coalition weights when doing numerical operations.
paired_shap_sampling	Logical. Whether to do paired sampling of coalitions.
prev_coal_samples	Character vector. A vector of previously sampled coalitions as characters. Each string contains a coalition and the feature indices in the coalition is separated by a space. For example, "1 5 8" is a coalition with features 1, 5, and 8.
prev_coal_samples_n_unique	Positive integer. The number of unique coalitions in prev_coal_samples. This is a separate argument to avoid recomputing the number unnecessarily.
n_samps_scale	Positive integer. Integer that scales the number of coalitions n_coalitions to sample as sampling is cheap, while checking for n_coalitions unique coalitions is expensive, thus we over sample the number of coalitions by a factor of n_samps_scale and determine when we have n_coalitions unique coalitions and only use the coalitions up to this point and throw away the remaining coalitions.
coal_feature_list	List. A list mapping each coalition to the features it contains.
approach0	Character vector. Contains the approach to be used for estimation of each coalition size. Same as approach in explain().
dt_valid_causal_coalitions	data.table. Only applicable for asymmetric Shapley values explanations, and is NULL for symmetric Shapley values. The data.table contains information about the coalitions that respects the causal ordering.

Value

A data.table with info about the coalitions to use

Author(s)

Nikolai Sellereite, Martin Jullum, Lars Henry Berge Olsen

Examples

## Not run: 
# All coalitions
x <- create_coalition_table(m = 3)
nrow(x) # Equals 2^3 = 8

# Subsample of coalitions
x <- shapr:::create_coalition_table(m = 10, exact = FALSE, n_coalitions = 1e2)

## End(Not run)

NorskRegnesentral/shapr documentation built on Feb. 11, 2025, 6:41 a.m.