Interpretability-Guided Bi-objective Optimization: Aligning Accuracy and Explainability

TL;DR

This paper presents IGBO, a bi-objective optimization framework that balances accuracy and interpretability by encoding feature hierarchies and using advanced attribution methods, with theoretical guarantees.

Contribution

It introduces a novel interpretability-guided bi-objective optimization method with DAG encoding, convergence proofs, and statistical guarantees for model interpretability.

Findings

Proposes a DAG-based feature importance hierarchy with statistical guarantees.

Develops a geometric projection for combining task and interpretability gradients.

Provides convergence proof to Pareto-stationary points.

Abstract

This paper introduces Interpretability-Guided Bi-objective Optimization (IGBO), a framework that trains interpretable models by incorporating structured domain knowledge via a bi-objective formulation. IGBO encodes feature importance hierarchies as a Directed Acyclic Graph (DAG) via Central Limit Theorem-based construction and uses Temporal Integrated Gradients (TIG) to measure feature importance. The framework employs a novel Relative Importance Score Hk(X, {\theta}) that quantifies the normalized cumulative attribution of each feature over time. We propose a geometric projection mapping P for combining task and interpretability gradients, and prove convergence to Pareto-stationary points. To address the Out-of-Distribution problem in TIG computation, we outline an Optimal Path Oracle architecture, which we leave for future work. Central Limit Theorem-based construction of the interpretability DAG provides statistical guarantees on acyclicity and transitivity, with an unconditional guarantee for the median threshold and conditional guarantees for higher confidence levels.