A Path to Simpler Models Starts With Noise

TL;DR

This paper investigates why simpler models can perform as well as complex models on noisy datasets, revealing that data noise and training choices expand the Rashomon set, with implications for model simplicity and interpretability.

Contribution

It introduces a mechanism linking data noise and training practices to large Rashomon ratios, and proposes pattern diversity as a measure of prediction variability in the Rashomon set.

Findings

Noisier datasets lead to larger Rashomon ratios.

Pattern diversity increases with label noise.

Simpler models can match complex models' performance on noisy data.

Abstract

The Rashomon set is the set of models that perform approximately equally well on a given dataset, and the Rashomon ratio is the fraction of all models in a given hypothesis space that are in the Rashomon set. Rashomon ratios are often large for tabular datasets in criminal justice, healthcare, lending, education, and in other areas, which has practical implications about whether simpler models can attain the same level of accuracy as more complex models. An open question is why Rashomon ratios often tend to be large. In this work, we propose and study a mechanism of the data generation process, coupled with choices usually made by the analyst during the learning process, that determines the size of the Rashomon ratio. Specifically, we demonstrate that noisier datasets lead to larger Rashomon ratios through the way that practitioners train models. Additionally, we introduce a measure called pattern diversity, which captures the average difference in predictions between distinct classification patterns in the Rashomon set, and motivate why it tends to increase with label noise. Our results explain a key aspect of why simpler models often tend to perform as well as black box models on complex, noisier datasets.