Multi-objective and multi-fidelity Bayesian optimization of laser-plasma acceleration

Beam parameter optimization in accelerators involves multiple, sometimes competing, objectives. Condensing these individual objectives into a single figure of merit unavoidably results bias towards particular outcomes, often an undesired way the absence prior knowledge. Finding optimal objective definition then requires operators to iterate over many possible weights and definitions, process that can take times longer than itself. A more versatile approach is multi-objective optimization, which establishes trade-off curve or Pareto front between Here we present first on Bayesian simulated laser-plasma accelerator. We find reaches comparable performance its single-objective counterparts while allowing for instant evaluation entirely new This dramatically reduces time required appropriate definitions problems. Additionally, our multi-objective, multi-fidelity method run by order magnitude. It does so dynamically choosing simulation resolution box size, requiring fewer slow expensive simulations as it learns about Pareto-optimal solutions from fast low-resolution runs. The techniques demonstrated this paper easily be translated different computational experimental use cases beyond accelerator optimization.