Event

Location: Room 2018, 1880 Pratt Drive, Corporate Research Center
Time: 2:30 PM

Speaker 1: Manikandan Soundarapandian

Title: FluCaster - a pervasive webapp for flu predictions & Postgres-XC speedup study and enhancements

Abstract: FluCaster is a disease surveillance and prediction tool that helps to visualize disease propagation in a spatio-temporal design. It utilizes data from Google Flu trends and epidemic simulation from Epifast. It also enables to run some predefined interventions like school closure, vaccination etc. Simdemics is the API engine that takes care of backend processing for the web app. I'll present my work on Simdemics briefly.
Postgres-XC is a multi-master write-scalable PostgreSQL cluster based on shared-nothing architecture. It is a collection of PostgreSQL database clusters which act as if the whole collection is a single database cluster. A table can either be distributed across the data nodes using round-robin, hash, modulo partitioning or can be replicated on all the data nodes. I'll go through some performance enhancements on which I've been working on.

Speaker 2: Sichao Wu

Title: A General Computational Framework for Experimental Design, Uncertainty Quantification, and Sensitivity Analysis

Abstract: Discrepancies often exist between computer simulation models and physical systems because of the endogenous and exogenous uncertainties lying in the models. Uncertainty quantification (UQ) and sensitivity analysis (SA) help to identify the most uncertain factors and provide insights to understand and control them. Both UQ and SA are essential ingredients of model verification and validation (V&V), which is a rigorous process to assure the model properly represents the reality that it is meant to capture. However, conducting UQ/SA studies is challenging for general domain experts without solid computing background since it requires extensive expertise in hardware, software, HPC, implementation, analysis, statistics, and the model itself. To address this, we propose a general computational framework for supporting computer experimental design and UQ/SA studies across various domains, simulation models, and running platforms. Within this framework, the domain experts will focus on the experimental design and analysis itself and the system will handle the formidable model execution and big data issues in a transparent way.