Mathematics of Efficient AI with Scientific Applications

Deep neural network (DNN) models in artificial intelligence (AI) have achieved remarkable success in various areas in the recent decade. Adopting AI becomes a timely choice to advance mathematics and scientific computing. On the other hand, the demands on computing resources and energy costs have gone up considerably, raising concerns on the sustainability of energy supply. Mathematics plays an essential role in achieving efficient AI (eAI) in terms of energy savings, computational complexity, data usage and accelerated inference.

The conference shall invite leading experts with diverse backgrounds to address the mathematics of eAI and its scientific applications. Specific topics include spikingneural networks and quantization, machine learning based reduced order models, efficient operator learning and foundation (transformer/diffusion) models, geometric/topological aspects of DNNs, eAI for inverse problems, medical imaging, fluid dynamics and material science.