June Presentation Night

Thank you to Think Big Analytics (http://thinkbig.teradata.com/) for sponsoring food and drink for this event.

Doors will open at 6PM for networking and food. Presentations will start at 6:30PM.

Jeffrey Byrne from STR will present on,

Face recognition has steadily improved in the past eight years, with modern algorithms recently demonstrating parity with human performance for near frontal or "mugshot" verification tasks. However, performance still degrades when faces exhibit large pose, illumination, expression and age changes. In this talk, we will discuss the challenges of these "corner faces" and summarize the state of the art in face recognition. We will specifically focus on the lessons learned using Spark for large scale machine learning for face recognition research at STR.

Anindita Mahapatra from Think Big will present on, Spark + GraphX to speed graph analytics for high tech manufacturing -

Many enterprise business problems are naturally modeled as a DAG and can be reduced to well understood graph analytic operations.

We have created a transitive closure to track product movement through a complex global value chain by sourcing product movement data from multiple legacy tracking systems. The transitive closure is invaluable in performing multiple analytic use cases which ultimately improve quality and profitability.

The Spark GraphX library and its use of Pregel-like processing framework was able to speed the transitive closure computation several magnitudes over Hive. We demonstrate how such an approach was integrated with the YARN Resource Manager and Hcatalog defined tables. We also describe how this system is used to continuously update the transitive closure on a lot latency 24x7 basis.