English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Several applications in the scientific simulation of physical systems can beformulated as control/optimization problems. The computational models for suchsystems generally contain hyperparameters, which control solution fidelity andcomputational expense. The tuning of these parameters is non-trivial and thegeneral approach is to manually `spot-check' for good combinations. This isbecause optimal hyperparameter configuration search becomes impractical whenthe parameter space is large and when they may vary dynamically. To addressthis issue, we present a framework based on deep reinforcement learning (RL) totrain a deep neural network agent that controls a model solve by varyingparameters dynamically. First, we validate our RL framework for the problem ofcontrolling chaos in chaotic systems by dynamically changing the parameters ofthe system. Subsequently, we illustrate the capabilities of our framework foraccelerating the convergence of a steady-state CFD solver by automaticallyadjusting the relaxation factors of discretized Navier-Stokes equations duringrun-time. The results indicate that the run-time control of the relaxationfactors by the learned policy leads to a significant reduction in the number ofiterations for convergence compared to the random selection of the relaxationfactors. Our results point to potential benefits for learning adaptivehyperparameter learning strategies across different geometries and boundaryconditions with implications for reduced computational campaign expenses.\footnote{Data and codes available at\url{https://github.com/Romit-Maulik/PAR-RL}}

Distributed deep reinforcement learning for simulation control