Search for: All records

AWP-ODC is a 4th-order finite difference code used for linear wave propagation, Iwan-type nonlinear dynamic rupture and wave propagation, and Strain Green Tensor simulation2. We have ported and verified the linear and topography version of AWP-ODC, with discontinuous mesh as well as topography, to HIP so that it can also run on AMD GPUs. The topography code achieved a 99.6% parallel efficiency on 4,096 nodes on Frontier, a Leadership Computing Facility at Oak Ridge National Laboratory. We have also implemented CUDA-aware features and on-the-fly GDR compression in the linear version of the ported HIP code. These enhancements significantly improve data transfer efficiency between GPUs, reducing communication overhead and boosting overall performance. We have also extended CUDA-aware features to the topography version and are actively working on incorporating GDR compression into this version as well. We see 154% benefits over IMPI in MVAPICH2-GDR with CUDA-aware support and on-the-fly compression for linear AWP-ODC on Lonestar-6 A100 nodes. Furthermore, we have successfully integrated a checkpointing feature into the nonlinear IWAN version of AWP-ODC, prepared for future extreme-scale simulation during Texascale Days of Frontera at TACC. 

In the last decade, there has been a surge in development and mainstream adoption of Artificial Intelligence (AI) systems that can generate textual image descriptions from images. However, only a few of these, such as Microsoft’s SeeingAI, are specifically tailored to needs of people who are blind screen reader users, and none of these have been brought to bear on the particular challenges faced by parents who desire image descriptions of children’s picture books. Such images have distinct qualities, but there exists no research to explore the current state of the art and opportunities to improve image-to-text AI systems for this problem domain. We conducted a content analysis of the image descriptions generated for a sample of 20 images selected from 17 recently published children’s picture books, using five AI systems: asticaVision, BLIP, SeeingAI, TapTapSee, and VertexAI. We found that descriptions varied widely in their accuracy and completeness, with only 13% meeting both criteria. Overall, our findings suggest a need for AI image-to-text generation systems that are trained on the types, contents, styles, and layouts characteristic of children’s picture book images, towards increased accessibility for blind parents. 

While active efforts are advancing medical artificial intelligence (AI) model development and clinical translation, safety issues of the AI models emerge, but little research has been done. We perform a study to investigate the behaviors of an AI diagnosis model under adversarial images generated by Generative Adversarial Network (GAN) models and to evaluate the effects on human experts when visually identifying potential adversarial images. Our GAN model makes intentional modifications to the diagnosis-sensitive contents of mammogram images in deep learning-based computer-aided diagnosis (CAD) of breast cancer. In our experiments the adversarial samples fool the AI-CAD model to output a wrong diagnosis on 69.1% of the cases that are initially correctly classified by the AI-CAD model. Five breast imaging radiologists visually identify 29%-71% of the adversarial samples. Our study suggests an imperative need for continuing research on medical AI model’s safety issues and for developing potential defensive solutions against adversarial attacks. 

The extent and ecological significance of intraspecific diversity within marine microbial populations is still poorly understood, and it remains unclear if such strain-level microdiversity will affect fitness and persistence in a rapidly changing ocean environment. In this study, we cultured 11 sympatric strains of the ubiquitous marine picocyanobacterium Synechococcus isolated from a Narragansett Bay (Rhode Island, USA) phytoplankton community thermal selection experiment. Despite all 11 isolates being highly similar (with average nucleotide identities of >99.9%, with 98.6-100% of the genome aligning), thermal performance curves revealed selection at warm and cool temperatures had subdivided the initial population into thermotypes with pronounced differences in maximum growth temperatures. Within the fine-scale genetic diversity that did exist within this population, the two divergent thermal ecotypes differed at a locus containing genes for the phycobilisome antenna complex. Our study demonstrates that present-day marine microbial populations can contain microdiversity in the form of cryptic but environmentally-relevant thermotypes that may increase their resilience to future rising temperatures.