Integrated Gradients for Model Explanations

Computes feature attributions using the Integrated Gradients method. This integrates gradients along the straight-line path from a baseline to each input.

integrated_gradients(
  f,
  X_train,
  X_new,
  baseline_method = "mean",
  n_steps = 50,
  h = 1e-05
)

Arguments

f

Function to explain (should accept matrix input, return vector output)

X_train

Training data (for computing baseline)

X_new

New observations to explain

baseline_method

Method for computing baseline: "mean", "median", or "zero"

n_steps

Number of steps for the integration path (default: 50)

h

Step size for finite differences (default: 1e-5)

Value

List with attributions, integrated gradients, baseline, predictions, and residual diagnostics

Examples

# integrated_gradients(f, X_train, X_new)