NAS-Bench-301 and the Case for Surrogate Benchmarks for Neural Architecture Search

Neural Architecture Search (NAS) is a logical next step in the automatic learning of representations, but the development of NAS methods is slowed by high computational demands. As a remedy, several tabular NAS benchmarks were proposed to simulate runs of NAS methods in seconds. However, all existing NAS benchmarks are limited to extremely small architectural spaces since they rely on exhaustive evaluations of the space. This leads to unrealistic results, such as a strong performance of local search and random search, that do not transfer to larger search spaces. To overcome this fundamental limitation, we propose NAS-Bench-301, the first model-based surrogate NAS benchmark, using a search space containing $10^{18}$ architectures, orders of magnitude larger than any previous NAS benchmark. We first motivate the benefits of using such a surrogate benchmark compared to a tabular one by smoothing out the noise stemming from the stochasticity of single SGD runs in a tabular benchmark. Then, we analyze our new dataset consisting of architecture evaluations and comprehensively evaluate various regression models as surrogates to demonstrate their capability to model the architecture space, also using deep ensembles to model uncertainty. Finally, we benchmark a wide range of NAS algorithms using NAS-Bench-301 allowing us to obtain comparable results to the true benchmark at a fraction of the cost.

Neural Ensemble Search for Performant and Calibrated Predictions

Ensembles of neural networks achieve superior performance compared to stand-alone networks not only in terms of accuracy on in-distribution data but also on data with distributional shift alongside improved uncertainty calibration. Diversity among networks in an ensemble is believed to be key for building strong ensembles, but typical approaches only ensemble different weight vectors of a fixed architecture. Instead, we investigate neural architecture search (NAS) for explicitly constructing ensembles to exploit diversity among networks of varying architectures and to achieve robustness against distributional shift. By directly optimizing ensemble performance, our methods implicitly encourage diversity among networks, without the need to explicitly define diversity. We find that the resulting ensembles are more diverse compared to ensembles composed of a fixed architecture and are therefore also more powerful. We show significant improvements in ensemble performance on image classification tasks both for in-distribution data and during distributional shift with better uncertainty calibration.

NAS-Bench-1Shot1: Benchmarking and Dissecting One-shot Neural Architecture Search

One-shot neural architecture search (NAS) has played a crucial role in making NAS methods computationally feasible in practice. Nevertheless, there is still a lack of understanding on how these weight-sharing algorithms exactly work due to the many factors controlling the dynamics of the process. In order to allow a scientific study of these components, we introduce a general framework for one-shot NAS that can be instantiated to many recently-introduced variants and introduce a general benchmarking framework that draws on the recent large-scale tabular benchmark NAS-Bench-101 for cheap anytime evaluations of one-shot NAS methods. To showcase the framework, we compare several state-of-the-art one-shot NAS methods, examine how sensitive they are to their hyperparameters and how they can be improved by tuning their hyperparameters, and compare their performance to that of blackbox optimizers for NAS-Bench-101.

Understanding and Robustifying Differentiable Architecture Search

Differentiable Architecture Search (DARTS) has attracted a lot of attention due to its simplicity and small search costs achieved by a continuous relaxation and an approximation of the resulting bi-level optimization problem. However, DARTS does not work robustly for new problems: we identify a wide range of search spaces for which DARTS yields degenerate architectures with very poor test performance. We study this failure mode and show that, while DARTS successfully minimizes validation loss, the found solutions generalize poorly when they coincide with high validation loss curvature in the space of architectures. We show that by adding one of various types of regularization we can robustify DARTS to find solutions with smaller Hessian spectrum and with better generalization properties. Based on these observations we propose several simple variations of DARTS that perform substantially more robustly in practice. Our observations are robust across five search spaces on three image classification tasks and also hold for the very different domains of disparity estimation (a dense regression task) and language modelling. We provide our implementation and scripts to facilitate reproducibility.

AutoDispNet: Improving Disparity Estimation with AutoML

Much research work in computer vision is being spent on optimizing existing network architectures to obtain a few more percentage points on benchmarks. Recent AutoML approaches promise to relieve us from this effort. However, they are mainly designed for comparatively small-scale classification tasks. In this work, we show how to use and extend existing AutoML techniques to efficiently optimize large-scale U-Net-like encoder-decoder architectures. In particular, we leverage gradient-based neural architecture search and Bayesian optimization for hyperparameter search. The resulting optimization does not require a large company-scale compute cluster. We show results on disparity estimation that clearly outperform the manually optimized baseline and reach state-of-the-art performance.

Towards Automated Deep Learning: Efficient Joint Neural Architecture and Hyperparameter Search

While existing work on neural architecture search (NAS) tunes hyperparameters in a separate post-processing step, we demonstrate that architectural choices and other hyperparameter settings interact in a way that can render this separation suboptimal. Likewise, we demonstrate that the common practice of using very few epochs during the main NAS and much larger numbers of epochs during a post-processing step is inefficient due to little correlation in the relative rankings for these two training regimes. To combat both of these problems, we propose to use a recent combination of Bayesian optimization and Hyperband for efficient joint neural architecture and hyperparameter search.