R-Adaptive Mesh Optimization to Enhance Finite Element Basis Compression

Watch a technical seminar presentation exploring innovative approaches to reduce memory usage in finite element discretizations while maintaining computational performance. Dive into dictionary-based data compression schemes that leverage mesh structure redundancies, achieving over 99% memory reduction in specific scenarios. Explore an augmented Lagrangian sequential quadratic programming algorithm for r-adaptive mesh optimization, designed to enhance mesh redundancies when natural structure is absent. Follow along as Denis Ridzal from Sandia National Laboratories addresses the challenge of balancing exascale computing capabilities with limited system memory, presenting numerical results from large-scale applications that demonstrate the effectiveness of these methods in detecting, exploiting, and enhancing mesh structure.