**Danny Bickson**

(Carnegie Mellon University)

Wednesday 15th February 2012

11:00am

** **

B10 Seminar Room, Basement,

Alexandra House, 17 Queen Square, London, WC1N 3AR

**Large scale iterative computation using GraphLab**

As the amounts of collected data and computing power grows (multicore, GPUs, clusters,clouds), modern datasets no longer fit into one computing node. Efficient distributed/parallel algorithms for handling large scale data are required.

In this talk, I will describe my work

concerning both the theoretical and the practical aspects of parallel and distributed large scale__ __computing on massive datasets. I will cover some developed theory and the required methods for handling heavy-tailed stable distributions in linear channels, which naturally characterize many system workloads like network flows in a communication network. We show for the first time how to compute inference in closed-form in a linear channel with additive stable noise.

The GraphLab framework is a parallel programming abstraction targeted for sparse iterative

graph algorithms. In the last year I have developed a multicore matrix factorization library on top of Graphlab.

By utilizing efficient parallel computation using Graphlab and BlackLight supercomptuer, we won the 5th place in last year ACM KDD CUP challenge (track1), out of more than 1000 participants.

This is a joint work with: Carlos Guestrin, Joseph Gonzalez, Yucheng Low, Aapo Kyrola__ __(Carnegie Mellon University) and Joseph Hellerstein (UC Berkeley).

**Bio:**

Danny Bickson is a project scientist at the Machine Learning Department in Carnegie Mellon University, hosted by Prof. Carlos Guestrin (CMU) and Prof. Joseph Hellerstein (UC Berkeley). His most recent project, GraphLab, involves the design and implementation of a distributed programming abstraction that outperforms MapReduce, designed to support iterative and potentially asynchronous algorithms on big data. Previously he was a Research staff member in the Messaging Group in IBM Haifa Research Lab, Israel, 2008-2009. He received his Ph.D. at the Hebrew University of Jerusalem in 2008, advised by Prof. Danny Dolev (HUJI) and Prof. Dahlia Malkhi (Microsoft Research Silicon Valley). His research targets large scale distributed algorithms design and their deployment, spanning both the theoretical and applied aspects of large scale computing and applied machine learning.