Event
Location: Second Floor Conference Room (2018)
Date: 04/10/2015
Time: 2:30 PM

Speaker 1: Achla Marathe

Title: Comparison of Private vs. Public Interventions for Controlling Influenza Epidemics

Abstract: This research studies how self-motivated individuals might react in the midst of an epidemic as they witness some of their immediate contacts become ill. We measure the impact of individual behavior and compare it with the impact of similar policies designed by public health officials and implemented as public health directives. There are several important differences between these strategies, e.g. individualistic behavioral modifications are often based on local information whereas public intervention is based on global information. Individuals react quickly and apply interventions immediately once their personal threshold is crossed whereas public health officials take longer to assess the situation and identify the appropriate intervention targets, often resulting in delay in implementing interventions.  The goal is to understand how individualistic actions, based on personal knowledge and beliefs, and aimed at self-protection, fare in comparison to similar actions imposed by public policy makers who depend on private citizens’ compliance.

Speaker 2: Md Hasanuzzaman Bhuiyan

Title: Apache Giraph: A Distributed Graph Processing Framework

Abstract: Many libraries and tools have been developed in recent years for processing large graphs. Among them Pregel, Giraph, GraphLab, GraphX and Parallel Boost Graph Library (PBGL) provide parallel and distributed graph processing framework. Han et al. showed that Giraph is one of the easiest system to understand and code for, as well as demonstrates good all-around performance for large scale graph processing. Giraph is an open source implementation of Pregel, a graph processing system developed and owned by Google, and inspired by Bulk Synchronous Parallel (BSP) model. Giraph is designed on top of Hadoop infrastructure with an efficient message passing model that makes it well-suited for graph algorithms. It scales up to billions of vertices and trillions of edges. It also offers fault tolerance that many graph libraries do not provide. A high level API hides the low level implementation details such as processor synchronization, coordination, communication and termination, from the users and thus makes it intuitive and easy to program. In this talk, I will give a high level overview of Giraph including the computation model, some optimization techniques and other features of Giraph.