GAMI-Net: An explainable neural network based on generalized additive models with structured interactions

PATTERN RECOGNITION

卷:120

Article

Yang, Zebin¹  Zhang, Aijun²  Sudjianto, Agus² 

[1] Univ Hong Kong, Dept Stat & Actuarial Sci, Pokfulam Rd, Hong Kong, Peoples R China

[2] Wells Fargo, Corp Model Risk, Sioux Falls, SD 57117 USA

关键词: Explainable neural network;    Generalized additive model;    Pairwise interaction;    Interpretability constraints;   

DOI  :  10.1016/j.patcog.2021.108192

来源: Elsevier

【 摘 要 】

The lack of interpretability is an inevitable problem when using neural network models in real applications. In this paper, an explainable neural network based on generalized additive models with structured interactions (GAMI-Net) is proposed to pursue a good balance between prediction accuracy and model interpretability. GAMI-Net is a disentangled feedforward network with multiple additive subnetworks; each subnetwork consists of multiple hidden layers and is designed for capturing one main effect or one pair wise interaction. Three interpretability aspects are further considered, including a) sparsity, to select the most significant effects for parsimonious representations; b) heredity, a pairwise interaction could only be included when at least one of its parent main effects exists; and c) marginal clarity, to make main effects and pairwise interactions mutually distinguishable. An adaptive training algorithm is developed, where main effects are first trained and then pairwise interactions are fitted to the residuals. Numerical experiments on both synthetic functions and real-world datasets show that the proposed model enjoys superior interpretability and it maintains competitive prediction accuracy in comparison to the explainable boosting machine and other classic machine learning models. (c) 2021 Elsevier Ltd. All rights reserved.

【 授权许可】

Free