M

Efficient global optimization of reactive force‐field parameters

“…Background information: OGOLEM OGOLEM is an object-oriented, easily extensible, platformindependent global optimization framework based on EAs, especially in the realization of GAs. [40,41] It combines threadlevel and MPI-level parallelism to achieve high scalability on shared memory as well as distributed memory architectures. The OGOLEM framework embodies our accumulated knowledge on nondeterministic global optimization in general and on EA s in particular [42,43] for various applications: cluster structures, [44][45][46][47][48][49][50][51][52][53][54] protein folding, [55] potential fitting, [34,35,[56][57][58][59][60] molecular design, [61] and abstract benchmarks.…”

“…The OGOLEM framework embodies our accumulated knowledge on nondeterministic global optimization in general and on EA s in particular [42,43] for various applications: cluster structures, [44][45][46][47][48][49][50][51][52][53][54] protein folding, [55] potential fitting, [34,35,[56][57][58][59][60] molecular design, [61] and abstract benchmarks. [62] EAs [19] borrow nomenclature from natural selection and evolution processes. To treat manifold optimization problems in a problem-independent manner, the problem specific system information, that is, everything that is defined as (indirect) input to the optimization function, is encoded as a genotype, a possible solution candidate is called an individual and the set of all individuals (and therefore their genotypes) present at a certain point in time is dubbed the genetic pool.…”

“…These scaling characteristics have already been observed in previous OGOLEM applications to different optimization tasks, for example with cluster structure optimization, [40,59] parameter fitting to traditional force fields [59] and abstract benchmarks. [62] Therefore, it can be taken as an intrinsic feature of the OGOLEM architecture.…”

“…In this article, we present further progress in algorithms and implementation to our earlier work on a single-objective GA optimization framework for ReaxFF. [34,35] We combine SPUREMD, [39] an advanced implementation of ReaxFF, with OGOLEM, [40,41] an advanced general evolutionary algorithm (EA) optimization suite. We show that the resulting framework produces results of at least the same quality as with our previous setup, [34,35] but in significantly shorter real times, offers better scalability and provides better user support and accessibility.…”

ParAMS: Parameter Optimization for Atomistic and Molecular Simulations

“…The main reason for this being a lack of generalization and transparency: Training data often comes in a variety of formats, optimizers expect a different input all together and the format in which parameters are stored is specific to each method. [35][36][37][38] The combination of these oftentimes results in works that can be hardly comprehended and reproduced by third parties. In an effort to address the above issues, we introduce the ParAMS scripting package for Python.…”