Model Validation Using Mutated Training Labels: An Exploratory Study

TL;DR

This paper proposes Mutation Validation, a novel model validation method that mutates training labels to assess model fit without using validation or test sets, showing high accuracy and stability across various algorithms and datasets.

Contribution

It introduces Mutation Validation, a new validation approach that relies on mutated training labels and metamorphic relations, avoiding traditional validation sets.

Findings

MV achieves 92% model selection accuracy

MV outperforms traditional validation in hyperparameter stability

MV is effective across multiple algorithms and datasets

Abstract

We introduce an exploratory study on Mutation Validation (MV), a model validation method using mutated training labels for supervised learning. MV mutates training data labels, retrains the model against the mutated data, then uses the metamorphic relation that captures the consequent training performance changes to assess model fit. It does not use a validation set or test set. The intuition underpinning MV is that overfitting models tend to fit noise in the training data. We explore 8 different learning algorithms, 18 datasets, and 5 types of hyperparameter tuning tasks. Our results demonstrate that MV is accurate in model selection: the model recommendation hit rate is 92\% for MV and less than 60\% for out-of-sample-validation. MV also provides more stable hyperparameter tuning results than out-of-sample-validation across different runs.