High-performance Computing

Thursday – October 31st

Session chair:

3:20 – 3:30

Session Intro and Welcome

Paul Hall

Senior Research Software Engineer, CCV

3:30 – 3:50

Physics Informed Neural Network (PINN): A data parallel GPU based approach

Raj Shukla

Research Software Engineer, CCV

Physical phenomena such as fluid flow, wave propagation, heat flow, quantum-mechanical system etc. are defined by partial differential equations (PDEs). To simulate these phenomena its essential to solve these PDEs using numerical discretization. Numerical discretization is very complex and suffers the curse of dimensionality and meshes. Thanks to Physics Informed Neural Network (PINN) for circumventing these problems. In this work, we show the efficiency of data parallel model, implemented over GPUs, to overcome the computational complexity of PINN models.

3:50 – 4:10

Demystifying the Gut Microbiome with Metagenomics

Swathi Penumutchu

Ph.D. Candidate, Pathobiology, MMI Department, Brown University

Using metagenomic pipelines to study fiber prebiotics and their potential to alleviate antibiotic-induced microbiome dysbiosis.

4:10 – 4:30

All the colors we cannot see: probing plant physiology with imaging spectroscopy

Loren Albert

Voss Postdoctoral Research Associate Institute at Brown for Environment and Society

Plants feed the world, drive the carbon cycle, and create habitat, but traditional methods of studying plant biology are slow and low throughput. Imaging spectroscopy—in which each pixel of an image acquires a spectrum instead of just three bands of RGB— may advance our capacity to study plant physiology and function. However, imaging spectroscopy presents new challenges in wrangling and processing big data.

4:30 – 4:50

Data Processing in the LUX Experiment

Will Taylor

Brown University/LUX Collaboration

The LUX detector was a dark matter search experiment which operated at the 4850 ft level of the Sanford Underground Research Facility (SURF). The experiment had a total operation time exceeding 3 years, during which time data was continuous transferred to Brown and processed in real time using the Oscar cluster. In total, LUX collected more than 500 TB of data, all of which was processed and reduced to analysis-ready features on Oscar. This talk will describe the details of the data handling and the strategies used to implement the processing scheme.

4:50 – 5:00

High-performance Computing Resources at Brown and Beyond.

Helen Kershaw

Lead Research Software Engineer, CCV

A summary of HPC resources available at Brown and beyond for research, educational and collaborative needs. Machines, workshops, courses, consulting and more.