Genetic Generalized Additive Models
arXiv:2602.15877v1 Announce Type: cross Abstract: Generalized Additive Models (GAMs) balance predictive accuracy and interpretability, but manually configuring their structure is challenging. We propose using the multi-objective genetic algorithm NSGA-II to automatically optimize GAMs, jointly minimizing prediction error (RMSE) and a Complexity Penalty that captures sparsity, smoothness, and uncertainty. Experiments on the California Housing dataset show that NSGA-II discovers GAMs that outperform baseline LinearGAMs in accuracy or match performance with substantially lower complexity. The resulting models are simpler, smoother, and exhibit narrower confidence intervals, enhancing interpretability. This framework provides a general approach for automated optimization of transparent, high-performing models. The code can be found at https://github.com/KaaustaaubShankar/GeneticAdditiveModels.
arXiv:2602.15877v1 Announce Type: cross Abstract: Generalized Additive Models (GAMs) balance predictive accuracy and interpretability, but manually configuring their structure is challenging. We propose using the multi-objective genetic algorithm NSGA-II to automatically optimize GAMs, jointly minimizing prediction error (RMSE) and a Complexity Penalty that captures sparsity, smoothness, and uncertainty. Experiments on the California Housing dataset show that NSGA-II discovers GAMs that outperform baseline LinearGAMs in accuracy or match performance with substantially lower complexity. The resulting models are simpler, smoother, and exhibit narrower confidence intervals, enhancing interpretability. This framework provides a general approach for automated optimization of transparent, high-performing models. The code can be found at https://github.com/KaaustaaubShankar/GeneticAdditiveModels.
Executive Summary
This article proposes a novel approach to optimize Generalized Additive Models (GAMs) using a multi-objective genetic algorithm, NSGA-II. The method jointly minimizes prediction error and a complexity penalty to achieve a balance between accuracy and interpretability. The authors demonstrate the effectiveness of their approach on the California Housing dataset, where the optimized GAMs outperform baseline LinearGAMs in accuracy or match performance with lower complexity. The resulting models are simpler, smoother, and exhibit narrower confidence intervals, enhancing interpretability. This framework provides a general approach for automated optimization of transparent, high-performing models.
Key Points
- ▸ The article proposes a novel approach to optimize GAMs using NSGA-II
- ▸ The method jointly minimizes prediction error and a complexity penalty
- ▸ The approach achieves a balance between accuracy and interpretability
- ▸ The optimized GAMs outperform baseline LinearGAMs on the California Housing dataset
Merits
Strength in Methodology
The use of NSGA-II to optimize GAMs is a novel and effective approach, allowing for automated optimization of transparent, high-performing models.
Strength in Results
The experimental results demonstrate the effectiveness of the approach, with optimized GAMs outperforming baseline LinearGAMs in accuracy or matching performance with lower complexity.
Demerits
Limitation in Generalizability
The approach is demonstrated on a single dataset, and its generalizability to other datasets and domains is unclear.
Limitation in Interpretability
While the optimized models are simpler and smoother, the complexity penalty may not fully capture the underlying structure of the data, potentially leading to over-simplification.
Expert Commentary
The article presents a novel and effective approach to optimizing GAMs using NSGA-II. However, the approach is limited by its generalizability to other datasets and domains. Additionally, the complexity penalty may not fully capture the underlying structure of the data, potentially leading to over-simplification. Nevertheless, the approach has significant practical and policy implications, and it has the potential to automate the optimization of GAMs in a variety of applications.
Recommendations
- ✓ Further research is needed to investigate the generalizability of the approach to other datasets and domains.
- ✓ The complexity penalty should be re-examined to ensure that it fully captures the underlying structure of the data.
