The AMG Summit

A research workshop in the high Rocky Mountains and on the California coast for the advancement of algebraic multigrid methods and related methods.



The policy of the Summit is that the participants leave their polished talks and reticence behind in favor of burning questions, open problems, new ideas, and readiness to interact.

Topics for the 2023 meeting

  • MG Teasers from Steve
    • Is MG O(n)?
    • Is MG sensitive to round-off?
    • If V(1,1) & TG(0,1) have similar rates, is one better than the other?
    • Is CG worthwhile for V(0,1)? For V(1,0)?
    • Is SOR unstable as an implicit time-stepping solver?
    • What properties do you want for the coarse-level matrix?
  • AMG interpolation for non-pointwise smoothers
  • AMG for Helmholtz and AMG for normal equations
  • AMG for PDEs with high-order terms
  • Parallel-in-time for
    • Hyperbolic problems
    • Chaotic problems
    • Optimization
  • Neural networks as a class of new approximating functions and its application in numerical PDEs
  • Optimizing AMG cycle structures, genetic algorithms
  • Reduction-based AMG
  • Parallel-in-time multigrid preconditioning w/ and w/o virtual variables
  • Stability of Coarse Grid Projection, finding compatible R and P for nonsymmetric A
  • Approximate domain decomposition smoothers, compressed solvers, and discretization
  • Constraint vectors for AIR
  • Data-driven exterior calculus
  • Machine learning
  • Information theory
  • Quantum-inspired multigrid