Bridging the Gap between Sample-based and One-shot Neural Architecture Search with BONAS
This work addresses the efficiency-reliability trade-off in NAS for machine learning practitioners, offering an incremental improvement by bridging sample-based and one-shot approaches.
The paper tackles the computational cost of sample-based Neural Architecture Search (NAS) by proposing BONAS, a framework that accelerates it using weight-sharing among related architectures, achieving significant speed improvements while maintaining reliability compared to one-shot NAS methods.
Neural Architecture Search (NAS) has shown great potentials in finding better neural network designs. Sample-based NAS is the most reliable approach which aims at exploring the search space and evaluating the most promising architectures. However, it is computationally very costly. As a remedy, the one-shot approach has emerged as a popular technique for accelerating NAS using weight-sharing. However, due to the weight-sharing of vastly different networks, the one-shot approach is less reliable than the sample-based approach. In this work, we propose BONAS (Bayesian Optimized Neural Architecture Search), a sample-based NAS framework which is accelerated using weight-sharing to evaluate multiple related architectures simultaneously. Specifically, we apply Graph Convolutional Network predictor as a surrogate model for Bayesian Optimization to select multiple related candidate models in each iteration. We then apply weight-sharing to train multiple candidate models simultaneously. This approach not only accelerates the traditional sample-based approach significantly, but also keeps its reliability. This is because weight-sharing among related architectures are more reliable than those in the one-shot approach. Extensive experiments are conducted to verify the effectiveness of our method over many competing algorithms.