Deterministic and statistical parameter space sampling: an autonomous driver for scientific workflows

Scientific workflows are usually controlled by many parameters, and assigning near-optimal values to these is often critical for efficiently finding solutions to goal-oriented problems. Such problems are typically solved by running sophisticated simulation or data analysis workflows. We present a novel sampling framework which is integrated into the Process Coordinator (ProC) - the general purpose scientific workflow engine originally developed for the Planck Surveyor satellite mission. The framework supports the exploration of high-dimensional parameter spaces for function representation, optimization, or integration purposes. Complemented by one of several pluggable sampling algorithms, a Sampler Control Element (SCE) drives the exploration process in multiple cycles. The whole sampling framework has been tested with different sampling algorithm plug-ins, and is ready for use in astrophysical research.