How to address monotonicity for model risk management?

Paper: arXiv 2305.00799

Authors: Dangxing Chen, Weicheng Ye

Abstract

In this paper, we study the problem of establishing the accountability and fairness of transparent machine learning models through monotonicity. Although there have been numerous studies on individual monotonicity, pairwise monotonicity is often overlooked in the existing literature. This paper studies transparent neural networks in the presence of three types of monotonicity: individual monotonicity, weak pairwise monotonicity, and strong pairwise monotonicity. As a means of achieving monotonicity while maintaining transparency, we propose the monotonic groves of neural additive models. As a result of empirical examples, we demonstrate that monotonicity is often violated in practice and that monotonic groves of neural additive models are transparent, accountable, and fair.

Complexity vs Empirical Score

• Math Complexity: 5.5/10
• Empirical Rigor: 6.0/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces a novel approach (monotonic groves of neural additive models) to address both individual and pairwise monotonicity constraints in transparent ML models for risk management. It combines moderate mathematical formulation of monotonicity constraints with empirical validation demonstrating practical violations and model fairness.

Research Flowchart

  flowchart TD
    A[Research Goal: Address Model Risk via Individual & Pairwise Monotonicity] --> B[Data Inputs: Tabular Datasets for Fairness & Transparency]
    B --> C[Analyze Existing Models for Monotonicity Violations]
    C --> D[Develop Method: Monotonic Groves of Neural Additive Models]
    D --> E[Enforce 3 Monotonicity Types: Individual, Weak Pairwise, Strong Pairwise]
    E --> F[Key Outcome: Demonstrated Monotonicity, High Transparency, and Model Fairness]