Re: reworking welkin

From: Stefano Mazzocchi <>
Date: Tue, 07 Feb 2006 11:13:56 -0500

Vineet Sinha wrote:
> Stefano Mazzocchi wrote:
>> Arvind Venkataramani wrote:
>>> ...
>>> 1. What are you people using welkin for? What kind of data?
>> I use it for a few things:
>> ...
>> 2) explore a subgraph of a bigger graph when I don't know what it is
>> (normally when I need to draw equivalences)
>> 3) explore a subgraph to understand graph-theoretical properties of
>> it (how connected it is, how many 'clusters' there are) and so on
> Stefano,
> You mention exploring subgraphs a number of times, when you work with
> subgraphs how do you obtain the subgraphs? Also, are there any
> particular types of subgraph queries that you use?

When I say 'subgraph' I mean a subset of the nodes that were loaded by
welkin. Welkin allows you to 'hide' nodes or links by simply interacting
with the UI, for example by clicking on the checkboxes on the tree on
the left, or by making rectangular selection on the degree distributions.

> Simply put there seem to be two parts to an information visualization
> task: performing queries to obtain the data, and the the actual display
> of the data. I am curious as to the types of queries that you perform,
> so that we can have a simple interface to put such data into various
> output formats.

We are talking about two things here:

  1) how to create a subgraph out of a very big graph

  2) how to explore such subgraphs into smaller sub-subgraphs that are
more meaningful to visualize (there are no more than 5/10 visual objects
a person can focus on a given time, but much more if you allow them to
interact with them)

One vision is to merge Welkin with Longwell, so that Longwell performs
#1 and welkin performs #2.

Another vision is to perform precalculation of graph clusters on the big
dataset and emerge another graph (not a subgraph!) that can be used as a
"graph guide", sort of a map of the graph, for exploratory purposes.

There are several algorithms in the literature to emerge such 'graph
maps' (for example, Newmann-Girman) but they tend to be extremely
expensive computation-wise.

But I do plan to work on this in the future, as part of the inferencing
engine for Longwell.

Stefano Mazzocchi
Research Scientist                 Digital Libraries Research Group
Massachusetts Institute of Technology            location: E25-131C
77 Massachusetts Ave                   telephone: +1 (617) 253-1096
Cambridge, MA  02139-4307              email: stefanom at mit . edu
Received on Tue Feb 07 2006 - 16:13:01 EST

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT