Quantum Dynamic Optimization Algorithm for Neural Architecture Search on Image Classification

Deep neural networks have proven to be effective in solving computer vision and natural language processing problems. To fully leverage its power, manually designed network templates, i.e., Residual Networks, are introduced deal with various tasks. These hand-crafted rely on a large number of parameters, which both data-dependent laborious. On the other hand, architectures suitable for specific tasks also grown exponentially their size topology, prohibits brute force search. address these challenges, this paper proposes quantum dynamic optimization algorithm find optimal structure candidate using Quantum Dynamic Neural Architecture Search (QDNAS). Specifically, proposed dynamics is used search meaningful dedicated rules express explore space. The treats iterative evolution process over time as process. tunneling effect potential barrier estimation mechanics can effectively promote global optimum. Extensive experiments four benchmarks demonstrate effectiveness QDNAS, consistently better than all baseline methods image classification Furthermore, an in-depth analysis conducted searchable that provide inspiration design networks.