Left Field: Graph Analytics and Big Data

Dear LPers,

Welcome to the first instance of a new column that we would like to launch. The title of this column is Left Field (a shorthand for “Out of Left Field”, an american expression to indicate something odd or unexpected – see http://en.wikipedia.org/wiki/Out_of_left_field). The objective of the Left Field column is to bring to the attention of the LP community articles, events, tools and anything else that does not properly belong to LP but where we feel that LP could have an important role.

We intend to publish at least one Left Field column for each issue of the newsletter – and that’s where you come into the picture: we need your perspectives and ideas. Are you working in other areas outside of LP and you see something interesting? Are you reviewing/reading a paper that has potential? Does your student discover an interesting article? Please send us a pointer along with a couple of sentences describing why you feel this could be of interest to the LP community. Not only, the more distant is from LP, the better. E.g., it is fairly obvious that a paper in functional programming or mathematical logic might have a relevance to LP – but it might be less obvious (and, thus, more exciting) to discover opportunities for LP in an article dealing with computer architectures or digital humanities. So… be bold!

We are looking forward your contributions!

Agostino & Enrico

Graph Analytics and Big Data

There is no doubt that the issue of “Big Data” is taking  a predominant role in the research community (at least here in the US). Dealing with large, complex, and heterogeneous data sets is complex, making sense out of them is a true challenge. While a significant effort is being invested in the traditional data mining approaches to extract patterns and rules from large data sets, there is a growing movement that is pushing towards big data analytics in the context of graph structured data sets. This is no surprise, considering the push of Google towards its Knowledge Graph, and the booming availability of RDF-based triple stores in several domains (e.g., in the life sciences).

We ran into this interesting workshop that took place last November, focuses specifically on the issue of Graph Analytics and Big Data. It was a BoaF session at the last Supercomputing conference. This seems like a very interesting domain for LP. LP (both in terms of ASP as well as CLP) has a significant advantage when it comes to processing graph based structures (especially when processing is driven by an underlying semantics of the graph, e.g., derived from an ontology). Yet, I suspect that the size of data we are dealing with are still out of the reach of most LP systems (that work predominantly with in-memory algorithms). This could provide an interesting challenge to LP in terms of exploring how inference algorithms can be adapted to become memory-aware (or memory-oblivious) and support the processing of large graph structures.

Pointer: http://www.graphanalysis.org/workshop2012-SC12.html

Title: Graph Analytics in Big Data (SC12 Birds-of-a-Feather session)

Contributed by: Enrico Pontelli and Agostino Dovier