Hi Andy,
Unfortunately, I was not at the meetup either; I'm located on the opposite
side of north america. I have, however been quite intrigued by talks and
lectures on TDA in the past, perhaps from a more academic point of view.
My interests are (broadly) in algebraic topology and learning, edging
towards computational topology; topological methods in ML are a (somewhat)
new interest, but I have a decent enough background in the pre-requisite
subjects to start coding some things up, I think. For fairly natural
reasons I would like to see more topological/geometric forms of
unsupervised learning in Python.
It is likely that I will implement some of these capabilities in Python
building on top of scikit-learn; I suppose the thing to do would be to make
a note of it on this mailing list once I have a decent working model and
see if there is any interest in folding it into the master branch or
otherwise make it into its own project. And yes, it may be a good idea for
me to give a few convincing examples or otherwise perhaps write a quick
whitepaper :-)
Some examples that leap to mind at the moment:
1) http://arxiv.org/abs/1307.1201 [R. Budney, W. Sethares. Topology of
Musical Data. ] Applications of persistent homology to discover topological
structures corresponding to musical structures;
2) http://www.math.upenn.edu/~ghrist/preprints/barcodes.pdf [R. Ghrist,
Persistent Topology of Data. [PDF] A survey article on topological methods;
also gives an overview of 3).
3) http://redwood.berkeley.edu/vs265/carlsson-ijcv08.pdf [G. Carlsson, T.
Ishkhanov, V. de Silva, A. Zomorodian. On the Local Behavior of Spaces of
Natural Images.] [PDF] Discovery of submanifold structure on a
higher-dimensional space of images.
Of course, these examples are somewhat specific. I'll be sure to make a
note if I manage to personally find a handful of applications that have
general/flexible enough appeal.
Anyway, it's purely speculative at this point, but it might very well be
interesting.
Cheers,
Paul
Message: 2
> Date: Thu, 24 Apr 2014 21:08:07 -0700
> From: Andy <[email protected]>
> Subject: Re: [Scikit-learn-general] Topological Data Analysis in
> sklearn
> To: [email protected]
> Message-ID: <[email protected]>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Hi Paul.
> Did you by any chance just go to a meetup in SF ;)
> I don't think there are any plans because these algorithms are not
> widely used in the ml community.
> If you can show how awesome they are for general ML, maybe ;) [I wasn't
> at the talk]
>
> Cheers,
> Andy
>
------------------------------------------------------------------------------
"Accelerate Dev Cycles with Automated Cross-Browser Testing - For FREE
Instantly run your Selenium tests across 300+ browser/OS combos. Get
unparalleled scalability from the best Selenium testing platform available.
Simple to use. Nothing to install. Get started now for free."
http://p.sf.net/sfu/SauceLabs
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
