Machine learning algorithm validation with a limited sample size

TLDR

High‑dimensional datasets with few samples, common in neuroimaging and genomics, are valuable for biomarker discovery but can bias machine‑learning performance estimates. The study investigates whether inadequate validation methods cause the observed bias in machine‑learning performance on small samples. Simulations examined how K‑fold cross‑validation, nested CV, train/test splits, feature selection, dimensionality, hyper‑parameter space, and fold number affect bias in small‑sample settings. K‑fold CV yields strongly biased estimates even with 1,000 samples, whereas nested CV and train/test splits remain robust; feature selection on pooled data introduces more bias than parameter tuning, providing guidance for designing reliable testing methods with limited data.

Abstract

Advances in neuroimaging, genomic, motion tracking, eye-tracking and many other technology-based data collection methods have led to a torrent of high dimensional datasets, which commonly have a small number of samples because of the intrinsic high cost of data collection involving human participants. High dimensional data with a small number of samples is of critical importance for identifying biomarkers and conducting feasibility and pilot work, however it can lead to biased machine learning (ML) performance estimates. Our review of studies which have applied ML to predict autistic from non-autistic individuals showed that small sample size is associated with higher reported classification accuracy. Thus, we have investigated whether this bias could be caused by the use of validation methods which do not sufficiently control overfitting. Our simulations show that K-fold Cross-Validation (CV) produces strongly biased performance estimates with small sample sizes, and the bias is still evident with sample size of 1000. Nested CV and train/test split approaches produce robust and unbiased performance estimates regardless of sample size. We also show that feature selection if performed on pooled training and testing data is contributing to bias considerably more than parameter tuning. In addition, the contribution to bias by data dimensionality, hyper-parameter space and number of CV folds was explored, and validation methods were compared with discriminable data. The results suggest how to design robust testing methodologies when working with small datasets and how to interpret the results of other studies based on what validation method was used.