In order to address the scalability challenge within Neural Architecture Search (NAS), we speed up NAS training via dynamic hard example mining within a curriculum learning framework. By utilizing an autoencoder that enforces an image similarity embedding in latent space, we construct an efficient kd-tree structure to order images by furthest neighbour dissimilarity in a low-dimensional embedding. From a given query image from our subsample dataset, we can identify the most dissimilar image within the global dataset in logarithmic time. Via curriculum learning, we then dynamically re-formulate an unbiased subsample dataset for NAS optimisation, upon which the current NAS solution architecture performs poorly. We show that our DDS-NAS framework speeds up gradient-based NAS strategies by up to 27x without loss in performance. By maximising the contribution of each image sample during training, we reduce the duration of a NAS training cycle and the number of iterations required for convergence.
@article{arxiv.2506.14667,
title = {DDS-NAS: Dynamic Data Selection within Neural Architecture Search via On-line Hard Example Mining applied to Image Classification},
author = {Matt Poyser and Toby P. Breckon},
journal= {arXiv preprint arXiv:2506.14667},
year = {2025}
}
Comments
27 single-column pages, 8 figures, to be published in Pattern Recognition