Propose your ideas

[mbastian]

We are currently preparing the Gephi's Google Summer of Code 2010 application. Inspired by our Roadmap, we are writing a set of subjects for various skills and level of difficulties. They will be available tomorrow on our wiki. But you are free to propose any idea, they will be considered and discussed on this topic.
Think what is possible to do with network visualization, how it could be useful in your company processes or how it could reveals hiddens patterns in any complex data.

So please share your ideas about Gephi in general or directly about Google SOC proposals !

[mbastian]

It's official! Gephi Google Summer of Code 2010 proposals are here.

* Shader Engine
* Dynamic attributes and statistics
* Web-based network visualization engine with WebGL
* Force-Directed Edge Bundling
* Data Laboratory
* Graph Streaming API
* Direct Social Networks Import

Please add some comments or correct mistakes on the document directly. If you have ressources that you think could be useful for a proposal, please add it too.

[pbittner]

Hi,

For the thesis project of my master degree in software engineering, I am going to work on Visualization of source code (from April to August). As it is part of the project to write a working piece of software (plug-in or standalone), I would like to know whether a source code visualization plug-in for Gephi is a suitable idea as part of the GSOC.

I know about CodeMiner (Roadmap 1.0) that let you create a graph of a source code project using functional dependencies. But from my understanding, the current version only generates links between elements (classes, methods...). It does not calculate any software metrics (fan-in/fan-out, cyclomatic complexity, depth of conditional nesting, length of code...) that can be used to determine the complexity, maintainability or reliability of the project.

My idea of source code visualization plug-in for Gephi would be to let the user import a project (GEXF from CodeMiner or direct import with static analysis at the pre-processing stage if feasible) and let him visualize the graph of the project at different scales (classes, methods) with each of the software metrics available. The plug-in would take advantage of the graph visualization functionalities of Gephi (layouts, filtering, ranking, clustering...) and use them to give a new perspective of source code visualization to the user. It would help a developer at different stages of the development process (testing, maintenance) by identifying the more complex or less reliable elements for instance. The user could also select from the plug-in window the packages or classes that he wants to visualize and the graph would be automatically updated.

I have lots of ideas and would be glad to write a detailed GSOC proposal if it is a suitable approach.

Thanks a lot,

Paul

[admin]

Hi Paul,

I'm glad to see that CodeMiner is an inspiration for GSoC proposals, and I foster you to write a detailed one!

Note that the current CodeMiner plugin for Eclipse, which produces GEXF files, is a frozen project, and we don't have any developer to make it evolve. As it isn't enough documented, I would advise you to start another source code analyzer from scratch, perhaps by using NetBeans IDE functionnalities. You know that Gephi is based on NetBeans platform, so it would be fairly easy to include some parts like the Java parser inside a Gephi plugin, which would generate the graph directly without the need of creating a GEXF!

Cheers,
Seb

[sumitd]

Hi,

My name is Sumit Dubey and I am interested in "Data Laboratory" project proposal. I have already introduced myself at http://forum.gephi.org/viewtopic.php?f=9&t=80 and filled and sent the questionaire to gsoc@gephi.org. I just wanted to know some time when I can find the mentors on IRC or whereever so that I can discuss the project. I guess I am not hurrying too much .

Thank you very much,

Sumit Dubey

[kumarlavr]

Hi,
I am Kumar Lav, Fourth Year Undergraduate from IIT Bombay.
While checking out the code from the launchpad by Bazaar, I am having proxy related issues.
I went through the Bazaar site but it wasnt helpful enough to solve the problem.
So, if any of you had already figured out, please let me know about it.
@sumit : I think ur colz is also behind proxy. So if you got it solved please do let me know.

With Regards,
Kumar Lav

[cezar_1]

Hi,

I am Cezary, a student of computer science at the Cybernetics Faculty, Military University of Technology, Warsaw.

I would be interested to know what you think about a project idea I've got. It's related with social networks. I have participated in a project concerning epidemics spreading and, as a part of the project, we have implemented several social networks generators (Barabási–Albert model for instance). I think this would be an excellent idea to implement these algorithms in Gephi. I have seen them in your Roadmap, so I assume you express some kind of interest in this topic. Do you think it is a good proposal for the GSoC project?

Best,
Cezary

[admin]

Hi Cezary,

This is indeed a good idea, as many researchers need to evaluate their algorithms on generated networks. But as the GSoC will last during 3 months, you have to propose enough algorithms/related work to be fully occupied. So could you list the algorithms you think important and for what kind of use?

Cheers,
Seb

[cezar_1]

Hi,

There are many generators I could implement, but I think that ones listed below will be enough for the proposal:

First of all I could implement several "basic" generators producing graphs of specific topologies, let's say:
http://mathworld.wolfram.com/PathGraph.html
http://mathworld.wolfram.com/StarGraph.html
http://mathworld.wolfram.com/CycleGraph.html
http://mathworld.wolfram.com/WheelGraph.html
http://mathworld.wolfram.com/LadderGraph.html
http://mathworld.wolfram.com/GridGraph.html
http://mathworld.wolfram.com/HypercubeGraph.html
http://mathworld.wolfram.com/BarbellGraph.html
http://mathworld.wolfram.com/CompleteGraph.html
http://mathworld.wolfram.com/FruchtGraph.html

These are very simple algorithms for a 'warm-up'.

After that I would like to implement networks generators, I am thinking about these ones:

1. Erdős-Rényi model in two versions G(n, p) and G(n, m). I would provide several implementations, there are some interesting documents dealing with them:

http://www.math-inst.hu/~p_erdos/1960-10.pdf
http://www.inf.uni-konstanz.de/algo/pub ... lrn-05.pdf

2. A few versions of Small World generators:

a) Alpha and Beta models of Watts and Strogatz, some interesting material:

http://tam.cornell.edu/tam/cms/manage/u ... lworld.pdf
http://www.bsos.umd.edu/socy/alan/stats ... -Ghosh.pdf

b) Kleinberg model http://citeseerx.ist.psu.edu/viewdoc/do ... 1&type=pdf

3. Several generators of Barabási–Albert model:

a) Basic generator often seen in popular libraries, for instance NetworkX.

b) Generalized generator with probabilities of adding new edges and rewiring existing ones.

c) PLOD http://citeseerx.ist.psu.edu/viewdoc/do ... 1&type=pdf

d) Simplified models: A (uniform attachment) and B (no growth).

http://arxiv.org/pdf/cond-mat/9910332v1

I also think about possibility of inducing the generators on an existing set of vertices.

Finally if time remains, I would implement generators for some interesting graphs listed here http://mathworld.wolfram.com/topics/SimpleGraphs.html - as many as it is posible.

So, what do you think?

[mbastian]

Thanks for this detailed list of possible generators! Its really useful for researchers to a have a broad list of generators to test their models.

However I don't think it overtake other priorities. It's preferable to focus on framework progress and develop new APIs. Adding new generators particularly fits to plug-in development and will not allow the student to interact with the core development so much.

But this list of generators is really great, we will put this on the wiki.