Neural Architecture Search (NAS) has become a popular method for discovering effective model architectures, especially for target hardware. As such, NAS methods that find optimal architectures under constraints are essential. In our paper, we propose LayerNAS to address the challenge of multi-objective NAS by transforming it into a combinatorial optimization problem, which effectively constrains the search complexity to be polynomial. For a model architecture with $L$ layers, we perform layerwise-search for each layer, selecting from a set of search options $\mathbb{S}$. LayerNAS groups model candidates based on one objective, such as model size or latency, and searches for the optimal model based on another objective, thereby splitting the cost and reward elements of the search. This approach limits the search complexity to $ O(H \cdot |\mathbb{S}| \cdot L) $, where $H$ is a constant set in LayerNAS. Our experiments show that LayerNAS is able to consistently discover superior models across a variety of search spaces in comparison to strong baselines, including search spaces derived from NATS-Bench, MobileNetV2 and MobileNetV3.

Artificial Intelligence (AI) is knocking on our doors with the goal of completely changing the user experience of "search" in the virtual world. Given the massive changes taking place in the technological sphere, AI is evidently the next big thing for humanity. The buzz surrounding ChatGPT and other similar tools pretty much justifies the excitement as well as the eagerness to adopt artificial intelligence in our daily lives. Recently, Google also introduced its own AI chat service called Bard, which some regard as a prominent competitor of ChatGPT and Bing AI. However, these AI tools aren't perfect yet, but they are a significant step forward in the development of AI assistant search and learning tools.