CSE Doctoral Student Seminar: Anthony Cabrera and James Orr

Apr 28, 2017
12:30 p.m.
2 p.m.
Lopata Hall, Room 101

"DIBS: A Benchmarking Suite for Data Integration, Wrangling, and Cleansing"

Anthony Cabrera
Adviser: Roger Chamberlain

Analyzing big data is a task encountered across disciplines. Addressing the challenges inherent in dealing with big data necessitate solutions that cover its three defining properties: volume, variety, and velocity. However, what is less understood is the treatment of the data that must be completed even before any analysis can begin. Specifically, there is often a non-trivial amount of time and resources that are utilized to the end of retrieving and preprocessing big data. This problem, known collectively as data integration, is a term frequently used for the general problem of taking data in some initial form and transforming it into a desired form. Examples of this include the rearranging of fields, changing the form of expression of one or more fields, altering the boundary notation of records and/or fields, encrypting or decrypting records and/or fields, parsing non-record data and organizing it into a record-oriented form, etc. In this work, we present our progress in creating a benchmarking suite that characterizes a diverse set of data integration applications.

"Towards Adaptive Cyber-Physical Systems"

James Orr
Adviser: Chris Gill

Recent advances in parallel real-time theory and platforms have allowed for previously unachievable temporal and computational resolution in high-performance cyber-physical applications such as real-time hybrid structural testing. However, these approaches suffer from rigidity and pessimism of resource allocation due to a priori analysis that is static at run-time.

To address those limitations we propose a new approach to exploit parallel real-time elasticity of online resource tradeoffs between computational and temporal resolution of different parts of the system. In doing so, we aim to use the slack between the allotted worst-case and actual behavior of individual sub-systems adaptively, to improve precision in managing computational resolution, temporal resolution, or both. This ongoing work lays the groundwork for a concurrency platform architecture that can exploit parallel real-time resource elasticity dynamically, towards more adaptive cyber-physical systems.