III: Small: Neural Field Representations for Scientific Visualization

Deep learning for scientific visualization has quickly emerged over the past few years as a vibrant direction in visualization research. Solutions based on well-known analytic approaches (e.g., convolutional neural networks and generative adversarial networks) have been extensively utilized to solve a range of scientific visualization problems. Visualizations science are key to helping people understand complex results and identifying issues in the analyses. This research pursues a novel direction for scientific visualization. This project will drive core visualization research, promote new methods for scientific visualization, which will impact not only computing, but other fields of study. The team will integrate research into education through special lectures, class projects, and a new course to educate students at the intersection of machine learning and data visualization. The investigator will continue attracting and recruiting undergraduate and underrepresented students through well-established institutional outreach programs and organize a conference tutorial to nourish future artificial intelligence researchers and the workforce. The research team will develop novel solutions that significantly augment the ability to synthesize, manage, explore, and communicate complex scientific data and their visualization output. The core research tasks feature visualization synthesis from sparse rendering images and expedited neural representation of visualization images. These tasks and their further possibilities cover essential topics, from visualization generation to neural compression and reconstruction. In addition, the team will perform comprehensive evaluations using multilevel metrics to assess the solution's effectiveness. The project outcomes will include neural field representation techniques for visualization synthesis and communication, comprehensive evaluations demonstrating the superiority of the proposed solutions compared with baseline and state-of-the-art solutions, and software libraries to benefit the scientific visualization community. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.