Adaptive parallelism with RMI: Idle high-performance computing resources can be completely avoided.

Spenke, Florian; Balzer, Karsten; Frick, Sascha; Hartke, Bernd; Dieterich, Johannes M.

doi:10.26434/chemrxiv.5809848.v1

Cited by 1 publication

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The core of this framework is a generation-free Genetic Algorithm (GA), including options for problem-specific “phenotype” operators, ,, niching, , and further ingredients that increase optimization efficiency. Additionally, ogolem features advanced parallelization techniques that allow for failsafe computing distributed across strongly heterogeneous hardware and for highly adaptive parallel runs that exploit traditionally inaccessible scheduling gaps . Partially, existing ogolem algorithm concepts have been reused or adapted to the current problem.…”

Section: Implementation and Methodsmentioning

confidence: 99%

Globally Optimal Catalytic Fields – Inverse Design of Abstract Embeddings for Maximum Reaction Rate Acceleration

Dittner

Hartke

2018

J. Chem. Theory Comput.

Self Cite

View full text Add to dashboard Cite

The search for, and understanding of, good catalysts for chemical reactions is a central issue for chemists. Here, we present first steps toward developing a general computational framework to better support this task. This framework combines efficient, unbiased global optimization techniques with an abstract representation of the catalytic environment, to shrink the search space. To analyze the resulting catalytic embeddings, we employ dimensionality reduction and clustering techniques. This not only provides an inverse design approach to new catalytic embeddings but also illuminates the actual interactions behind catalytic effects. All this is illustrated here with a strictly electrostatic model for the environment and with two versions of a selected example reaction. We close with detailed discussions of future improvements of our framework.

show abstract