Large-Scale Graph Data Mining with MapReduce: the Bag of Algorithmic Tricks

Nima Sarshar, Ph.D.
Intuit

Agenda

6:30 pm Networking, Snacks
7:00 pm Announcements
7:10 pm Presentation and Q&A

Event Details

Many modern large-scale data mining problems are defined on graphs (think of People you May Know) , or have a graph representation (think of Collaborative filtering and it’s bi-partite graph representation). This makes Hadoop and the MapReduce framework natural candidates to tackle them. Some graph processing algorithms, e.g. global PageRank, can be ported into the MapReduce framework rather straightforwardly. Others require various degrees of combinatorial tricks. In this talk, we review several fundamental graph processing algorithms that require careful, and often beautiful, tricks to scale when dealing with very large graphs. These include enumerating triangles and rectangles (e.g., to find Friends in Common at scale) , creating induced latent networks and collaborative filtering on bi-partite graphs, Personalized PageRank and more. We will describe some of the applications of these algorithms at Intuit.

Speaker Bio

Nima is a Senior Data Scientist at Intuit. Before Intuit he was the co-founder and CTO of Haileo Inc, a Santa Clara based star up specializing in context-based video advertisement. Before that, he was an Associate Prof. of Software Engineering at the University of Regina, Canada.

Event page provided by ACM
https://www.sfbayacm.org/event/large-scale-graph-data-mining-mapreduce-bag-algorithmic-tricks