Speeding up NAS with Adaptive Subset Selection

Vishak Prasad C, Colin White, Sibasis Nayak, Paarth Jain, Aziz Shameem, Prateek Garg, Ganesh Ramakrishnan

Keywords: Neural Architecture Search, Subset Selection, Optimisation

Abstract: A majority of recent developments in neural architecture search (NAS) have been aimed at decreasing the computational cost of various techniques without affecting their final performance. Towards this goal, several low-fidelity and performance prediction methods have been considered, including those that train only on subsets of the training data. In this work, we present an adaptive subset selection approach to NAS and present it as complementary to state-of-the-art NAS approaches. We uncover a natural connection between one-shot NAS algorithms and adaptive subset selection and devise an algorithm that makes use of state-of-the-art techniques from both areas. We use these techniques to substantially reduce the runtime of DARTS-PT (a leading one-shot NAS algorithm), as well as BOHB and DEHB (leading multi-fidelity optimization algorithms), with minimal sacrifice to accuracy. In experiments, we find architectures on CIFAR-10 that give 5\% increase in performance over DARTS-PT while reducing the time required by more than 8 times. Our results are consistent across multiple datasets, and towards full reproducibility, we release all our code at \url{https://anonymous.4open.science/r/SubsetSelection_NAS-87B3}

Submission Checklist: Yes

Broader Impact Statement: Yes

Paper Availability And License: Yes

Code Of Conduct: Yes

Optional Meta-Data For Green-AutoML: This blue field is just for structuring purposes and cannot be filled.

Submission Number: 3