Workshop Scope

Scientific imaging and spatiotemporal datasets—from medical imaging, climate and weather, Earth systems, materials science, manufacturing, bioengineering, microscopy, cosmology, and nuclear fusion—are growing rapidly in spatial and temporal resolution, duration, modality, volume, and dimensionality. Extracting meaningful scientific information from these increasingly complex datasets requires computational capabilities that often exceed those of conventional algorithms and computing workflows.

Leadership-class high-performance computing (HPC) systems offer an opportunity to go beyond simply processing larger datasets or accelerating existing workflows. By scaling computational methods, algorithms, and systems, researchers can reconstruct finer-scale structures, resolve longer-term dynamics, integrate heterogeneous observations, quantify uncertainty, and analyze complex scientific phenomena at previously inaccessible scales. These capabilities can be advanced through a broad range of approaches, including artificial intelligence (AI), computational imaging, inverse problems, numerical algorithms, scientific modeling, data assimilation, and scalable data analysis.

Scale to See establishes an interdisciplinary forum for advancing the computational methods and HPC capabilities needed to transform large-scale scientific imaging and spatiotemporal data into new scientific understanding. The workshop welcomes both AI-based and non-AI approaches, with an emphasis on methods that exploit extreme-scale computing to overcome fundamental computational bottlenecks and enable new scientific capabilities. Of particular interest are approaches that combine algorithmic innovation, scientific knowledge, and HPC system design to achieve greater resolution, scale, fidelity, efficiency, and reliability.

Through contributions spanning algorithms, AI, HPC systems, and scientific applications, Scale to See aims to foster cross-disciplinary collaboration and advance a central vision: computational scale should do more than make existing science faster—it should make new science possible.

The workshop is guided by the central question:

How can scaling computational methods and HPC systems on imaging or spatiotemporal data enable us to see, infer, and understand scientific phenomena that were previously inaccessible?

More specifically, the workshop will explore the following questions:

• Scalable algorithms and computing systems: How can we design algorithms, software, and computing architectures that efficiently exploit leadership-class HPC systems for scientific imaging and spatiotemporal data analysis?

• AI and computational methods: How can AI, inverse-problem methods, numerical algorithms, and scientific modeling complement one another to tackle increasingly large and complex scientific problems?

• Scientific fidelity and uncertainty: How can physical principles, scientific knowledge, heterogeneous observations, and uncertainty quantification be incorporated into computational methods at extreme scale?

• From scale to scientific capability: How can increased spatial and temporal resolution, longer context, larger ensembles, and multimodal data integration enable analyses and inferences that were previously computationally infeasible?

• New scientific opportunities: What new observations, reconstructions, predictions, measurements, and discoveries become possible when computational methods are scaled beyond conventional limits?

                                                                     Topics of Interest

Scale to See welcomes contributions that advance the computational methods, algorithms, and computing systems needed to push the boundaries of imaging and high-dimensional spatiotemporal data analysis. We are particularly interested in work where scaling computational capability enables new forms of scientific observation, reconstruction, inference, and discovery that are difficult or infeasible with conventional approaches.

Topics span AI-based and non-AI methods, scientific algorithms, and HPC systems. We encourage contributions that address the challenges of scaling along multiple dimensions, including spatial resolution, temporal extent, model complexity, ensemble size, modality, and data dimensionality.

Topics include, but are not limited to:

Theme 1: Beyond Conventional Limits — Extreme-Scale Scientific Observation and Inference

• Ultra-high-resolution scientific imaging, reconstruction, and enhancement beyond conventional computational limits

• Large-scale computational imaging and inverse problems

• Global-scale, multiscale, and long-context spatiotemporal analysis

• Scientific sensing, observation, and reconstruction from sparse, incomplete, or heterogeneous measurements

• Multimodal data integration and fusion across instruments, simulations, and observational platforms

• Large-scale state estimation, scientific forecasting, and data assimilation

• High-dimensional scientific data analysis across spatial, temporal, and physical scales

Theme 2: Computational Methods That Turn Scale into Scientific Capability

• Scalable AI and scientific foundation models for imaging and spatiotemporal data

• Generative modeling, posterior sampling, and uncertainty-aware scientific inference

• Physics-informed AI, model-based learning, and hybrid scientific computing methods

• Scalable numerical algorithms, optimization, and solvers for imaging and inverse problems

• Self-supervised learning, scientific representation learning, and data-efficient methods

• Large-ensemble methods for uncertainty quantification, rare-event analysis, and probabilistic prediction

• Coupled AI, numerical modeling, and data assimilation methods

• New algorithmic formulations that exploit computational scale to improve scientific fidelity or enable previously inaccessible analyses

Theme 3: Extreme-Scale Algorithms, Systems, and Co-Design

• Parallel and distributed algorithms for scientific imaging and spatiotemporal data

• HPC–AI co-design and hardware–software co-design for scientific workloads

• Exascale computing methods for large-scale reconstruction, inference, and analysis

• Scalable training, inference, and execution of scientific models

• Memory-, communication-, and I/O-efficient algorithms for high-dimensional scientific data

• Scalable data management, workflow orchestration, and in situ or in-transit processing

• AI compilers, runtime systems, and programming models for scientific computing

• Performance modeling, scalability analysis, benchmarking, and reproducibility on leadership-class HPC systems

Theme 4: Scientific Capabilities and Discoveries Enabled by Scale

• Scientific applications demonstrating new capabilities enabled by increased spatial or temporal resolution, dimensionality, or computational scale

• High-fidelity scientific digital twins and multiscale reconstruction or simulation

• Long-horizon prediction and analysis of complex evolving systems

• Extreme-event characterization, rare-event discovery, and tail-risk quantification

• Real-time or near-real-time scientific inference and decision support

• Scalable analysis of experimental and simulation data to guide experiments, observations, and scientific workflows

• Case studies demonstrating scientific questions, structures, or phenomena that become accessible through advances in scalable algorithms and computing

Submission Format:

• Paper Length: 4–8 pages (excluding references)

• Format: SC Proceedings Template 

• The review process is double-blind, i.e. the names of the authors, reviewers, and area chairs are not revealed to each other. Papers must thus be properly anonymized before submission. 

• Domain conflicts: To avoid conflict of interest among the authors, reviewers and meta-reviewers, all co-author information and a complete and accurate list of domain conflicts must be properly entered in the SC submission site by the submission deadline. 

Paper Publication:

Accepted papers will be published in SC'26 proceedings as workshop papers.