On the Codesign of Scientific Experiments and Industrial Systems

Abstract

The optimization of large experiments in fundamental science, such as detectors for subnuclear physics at particle colliders, shares with the optimization of complex systems for industrial or societal applications the common issue of addressing the inter-relation between parameters describing the hardware used in data production and parameters used to analyse those data. While in many cases this coupling can be ignored -- when the problem can be successfully factored into simpler sub-tasks and the latter addressed serially -- there are situations in which that approach fails to converge to the absolute maximum of expected performance, as it results in a mis-alignment of the optimized hardware and software solutions. In this work we consider a few use cases of interest in fundamental science collected primarily from particle physics and related areas, and a pot-pourri of industrial and societal applications where the matter is similarly of relevance. We discuss the emergence of strong hardware-software coupling in some of those systems, as well as co-design procedures that may be deployed to identify the global maximum of their relevant utility functions. We observe how numerous opportunities exist to advance methods and tools for hardware-software co-design optimization, bridging fundamental science and industry through application- and challenge-driven projects, and shaping the future of scientific experiments and industrial systems.

Figures (53)

• Figure 1: Simple examples of different classes of optimization problems. For visual clarity, levels are scaled uniformly in $\log\left(1 + f(h, s)\right)$.
• Figure 2: Schematic of the result of optimization schemes on block-separable (left) and non-block-separable (center and right) problems. The elliptic contours identify points at same value of the utility. On the left, in a separable problem a hardware optimization (vertical arrow) followed by a software tuning (horizontal arrow) will always reach the highest value of the utility. At the center and on the right, in a non-separable problem the outcome of that modus operandi will depend on the initial software chosen for the hardware parameters scan -- here exemplified as four equispaced initial parameter values on the horizontal axis. On the right, gradient descent on the full space identifies the global extremum of the utility.
• Figure 3: Workflow of AI-assisted detector design ai-tracking-eic.
• Figure 4: Scheme of a simplified tracker made of 33 silicon strips (blue or green segments along $y$) organized into three layers (yellow) . The passage of a straight-going particle (red) may be recorded by the three green-coloured strips.
• Figure 10: Optimization schemes for both AIDO and BO methods. $\Theta$ represents the set of optimizable detector parameters, $E$ and $C$ are the true targets for particle energy and type, $E_{reco}$ and $C_{reco}$ their reconstructed quantities. While AIDO (top) uses a generative surrogate model---trained to emulate the reconstruction---from which a gradient can be obtained to optimize $\Theta$, BO (bottom) uses a Bayesian surrogate to map the loss space as a function of $\Theta$.

Theorems & Definitions (2)

• Definition 1
• Definition 2