Generalization Guarantees for Neural Architecture Search with Train-Validation Split

TL;DR

This paper analyzes the statistical properties of neural architecture search with train-validation splits, showing how validation metrics can guide generalization and proposing bounds and methods for effective NAS.

Contribution

It provides new theoretical insights into NAS generalization, bounds for gradient-based search, and connections to kernel and matrix learning methods.

Findings

Validation loss properties indicate true test loss.

Gradient descent finds optimal architecture even with zero training error.

Spectral methods can efficiently solve the outer NAS problem.

Abstract

Neural Architecture Search (NAS) is a popular method for automatically designing optimized architectures for high-performance deep learning. In this approach, it is common to use bilevel optimization where one optimizes the model weights over the training data (inner problem) and various hyperparameters such as the configuration of the architecture over the validation data (outer problem). This paper explores the statistical aspects of such problems with train-validation splits. In practice, the inner problem is often overparameterized and can easily achieve zero loss. Thus, a-priori it seems impossible to distinguish the right hyperparameters based on training loss alone which motivates a better understanding of the role of train-validation split. To this aim this work establishes the following results. (1) We show that refined properties of the validation loss such as risk and hyper-gradients are indicative of those of the true test loss. This reveals that the outer problem helps select the most generalizable model and prevent overfitting with a near-minimal validation sample size. This is established for continuous search spaces which are relevant for differentiable schemes. Extensions to transfer learning are developed in terms of the mismatch between training & validation distributions. (2) We establish generalization bounds for NAS problems with an emphasis on an activation search problem. When optimized with gradient-descent, we show that the train-validation procedure returns the best (model, architecture) pair even if all architectures can perfectly fit the training data to achieve zero error. (3) Finally, we highlight connections between NAS, multiple kernel learning, and low-rank matrix learning. The latter leads to novel insights where the solution of the outer problem can be accurately learned via efficient spectral methods to achieve near-minimal risk.