Accelerating Legacy Numerical Solvers by Non-intrusive Gradient-based Meta-solving

arXiv:2405.02952v1 Announce Type: new
Abstract: Scientific computing is an essential tool for scientific discovery and engineering design, and its computational cost is always a main concern in practice. To accelerate scientific computing, it is a promising approach to use machine learning (especially meta-learning) techniques for selecting hyperparameters of traditional numerical methods. There have been numerous proposals to this direction, but many of them require automatic-differentiable numerical methods. However, in reality, many practical applications still depend on well-established but non-automatic-differentiable legacy …

abstract arxiv computational computing cost cs.lg cs.na design discovery engineering engineering design gradient machine machine learning math.na meta meta-learning numerical practice scientific scientific discovery tool type