Engineering Papers⌕ Search

Engineering topics

Franklin, Michael J.

Publications and source records attributed to Franklin, Michael J..

DLHub: Simplifying publication, discovery, and use of machine learning models in science

Machine Learning (ML) has become a critical tool enabling new methods of analysis and driving deeper understanding of phenomena across scientific disciplines. There is a growing need for "learning systems" to support various phases in the ML lifecycle. While others have focused on supporting model development, training, and inference, few have focused on the unique challenges inherent in science, such as the need to publish and share models and to serve them on a range of available computing resources. In this paper, we present the Data and Learning Hub for science (DLHub), a learning system designed to support these use cases. Specifically, DLHub enables publication of models, with descriptive metadata, persistent identifiers, and flexible access control. It packages arbitrary models into portable servable containers, and enables low-latency, distributed serving of these models on heterogeneous compute resources. In this work, we show that DLHub supports low-latency model inference comparable to other model serving systems including TensorFlow Serving, SageMaker, and Clipper, and improved performance, by up to 95%, with batching and memoization enabled. We also show that DLHub can scale to concurrently serve models on 500 containers. Finally, we describe five case studies that highlight the use of DLHub for scientific applications.

97 MATHEMATICS AND COMPUTING↗

Kira: Processing Astronomy Imagery Using Big Data Technology

Scientific analyses commonly compose multiple single-process programs into a dataflow. An end-to-end dataflow of single-process programs is known as a many-task application. Typically, HPC tools are used to parallelize these analyses. In this work, we investigate an alternate approach that uses Apache Spark-a modern platform for data intensive computing-to parallelize many-task applications. We implement Kira, a flexible and distributed astronomy image processing toolkit, and its Source Extractor (Kira SE) application. Using Kira SE as a case study, we examine the programming flexibility, dataflow richness, scheduling capacity and performance of Apache Spark running on the Amazon EC2 cloud. By exploiting data locality, Kira SE achieves a 4.1× speedup over an equivalent C program when analyzing a 1TB dataset using 512 cores on the Amazon EC2 cloud. Furthermore, Kira SE on the Amazon EC2 cloud achieves a 1.8× speedup over the C program on the NERSC Edison supercomputer. A 128-core Amazon EC2 cloud deployment of Kira SE using Spark Streaming can achieve a second-scale latency with a sustained throughput of 800 MB/s. Our experience with Kira demonstrates that data intensive computing platforms like Apache Spark are a performant alternative for many-task scientific applications.

97 MATHEMATICS AND COMPUTING↗

Microsensor and transcriptomic signatures of oxygen depletion in biofilms associated with chronic wounds: Biofilms and oxygen

Polymicrobial biofilms have been implicated in delayed wound healing, although the mechanisms by which biofilms impair wound healing are poorly understood. Many species of bacteria produce exotoxins and exoenzymes that may inhibit healing. In addition, oxygen consumption by biofilms may impede wound healing. In this study, we used oxygen microsensors to measure oxygen transects through in vitro-cultured biofilms, biofilms formed in vivo in a diabetic (db/db) mouse model, and ex vivo human chronic wound specimens. The results show that oxygen levels within both euthanized and live mouse wounds had steep gradients that reached minima ranging from 19 to 61% oxygen partial pressure, compared to atmospheric oxygen levels. The oxygen gradients in the mouse wounds were similar to those observed for clinical isolates cultured in vitro and for human ex vivo scabs. No oxygen gradients were observed for heat-killed scabs, suggesting that active metabolism by the viable bacteria contributed to the reduced oxygen partial pressure of the wounds. To characterize the metabolic activities of the bacteria in the mouse wounds, we performed transcriptomics analyses of Pseudomonas aeruginosa biofilms associated with the db/db mice wounds using Affymetrix microarrays. The results demonstrated that the bacteria expressed genes for metabolic activities associated with cell growth. Interestingly, the transcriptome results indicated that the bacteria within the wounds also experienced oxygen-limitation stress. Among the bacterial genes that were expressed in vivo were genes associated with the Anr-mediated hypoxia-stress response. Other bacterial stress response genes highly expressed in vivo were genes associated with stationary-phase growth, osmotic stress, and RpoH-mediated heat shock stress. Overall, the results support the hypothesis that the metabolic activities of bacteria in biofilms act as oxygen sinks in chronic wounds and that the depletion of oxygen contributes to the detrimental impact of biofilms on wound healing.

bacterial biofilms↗