From Sequential Nodes to GPU Batches: Parallel Branch and Bound for Optimal $k$-Sparse GLMs
Researchers have developed a new CPU-GPU framework to accelerate optimization problems with discrete variables, which have historically been challenging for GPUs. This framework processes branch and bound nodes in batches on GPUs, overcoming issues of sequential processing and data movement. Experiments demonstrate significant speedups and the ability to collect the full Rashomon set for further statistical analysis. AI
IMPACT Enables faster and more comprehensive analysis of complex models, potentially improving downstream AI applications.