From Sorting Algorithms to Scalable Kernels: Bayesian Optimization in High-Dimensional Permutation Spaces
Researchers have developed a new framework for Bayesian Optimization (BO) in high-dimensional permutation spaces, addressing the limitations of current methods that struggle with scalability. The proposed approach utilizes kernel functions derived from sorting algorithms, introducing a novel 'Merge Kernel' based on merge sort. This kernel offers a compact representation with a complexity of \Theta(n\log n), outperforming the traditional Mallows kernel in both optimization performance and computational efficiency as dimensionality increases. The findings suggest this method can tackle complex problems like large-scale feature ordering and combinatorial neural architecture search. AI
IMPACT Enables more efficient optimization for complex AI tasks like neural architecture search.