Flexible Bayesian Nonlinear Model Configuration
This addresses the need for interpretable models in scientific research where complex relationships exist, offering a flexible alternative to neural networks or specific parametric functions, though it appears incremental as it builds on existing Bayesian and feature generation methods.
The paper tackles the problem of building interpretable nonlinear regression models by introducing a flexible Bayesian approach that hierarchically generates nonlinear features, similar to deep learning, and uses variable selection to find a small set of important features. It demonstrates predictive performance comparable to several machine learning algorithms in various applications.
Regression models are used in a wide range of applications providing a powerful scientific tool for researchers from different fields. Linear, or simple parametric, models are often not sufficient to describe complex relationships between input variables and a response. Such relationships can be better described through flexible approaches such as neural networks, but this results in less interpretable models and potential overfitting. Alternatively, specific parametric nonlinear functions can be used, but the specification of such functions is in general complicated. In this paper, we introduce a flexible approach for the construction and selection of highly flexible nonlinear parametric regression models. Nonlinear features are generated hierarchically, similarly to deep learning, but have additional flexibility on the possible types of features to be considered. This flexibility, combined with variable selection, allows us to find a small set of important features and thereby more interpretable models. Within the space of possible functions, a Bayesian approach, introducing priors for functions based on their complexity, is considered. A genetically modified mode jumping Markov chain Monte Carlo algorithm is adopted to perform Bayesian inference and estimate posterior probabilities for model averaging. In various applications, we illustrate how our approach is used to obtain meaningful nonlinear models. Additionally, we compare its predictive performance with several machine learning algorithms.