Minutes from Nov 19 Telcon

Thanks for scribing, Gregg!

Minutes from Nov 19 Telcon

[11:59] <Ashok> meeting:Property Graphs CG
[11:59] <Ashok> chair: Ashok
[12:00] <Ashok> data:Nov 19, 2013
[12:04] <Ashok> present: Ashok, Zhe, Gregg, Kelvin, Patrick
[12:06] <gkellogg> scribe: gkellogg
[12:06] <Ashok> Minutes of last meeting:http://lists.w3.org/Archives/Public/public-propertygraphs/2013Nov/0009.html
[12:06] <zwu2> looks good to me
[12:07] <gkellogg> RESOLVED: approved minutes from last week
[12:07] <gkellogg> Ashok: Next week is Thanksgiving, are people going to be around, or should we cancel?
[12:07] <zwu2> it is probably the best to skip it
[12:08] <gkellogg> Ashok: I'm also inclined to skip ...I'd like to get through 1-2 of the use cases. I'll put them on the Wiki, and we can start working on those.
[12:09] <gkellogg> Ashok: Next tuesday's call is canceled; the next call will be 3 December
[12:09] <gkellogg> Topic: Use Cases
[12:09] <gkellogg> Ashok: 2 use cases, one from Michael, and one from Patrick.
[12:09] <gkellogg> ... Patrick's is from a different area.
[12:10] <gkellogg> Patrick: This was out of Bio-Informatics. I've been doing a fair amount of research on graphs from a semantic perspective.
[12:10] <gkellogg> ... I encountered an article on protien-protien interactions.
[12:10] <gkellogg> ... The article suggested there were a number of advantages for using PGs.
[12:11] <gkellogg> ... Finding the shortest-path is an essential algorithm.
[12:11] <gkellogg> ... Scoring paths between proteins; that is the path that is most significant.
[12:12] <gkellogg> ... They're interested in the most likely paths that they may be connected, based on determining the strengths.
[12:12] <gkellogg> ... By picking out the path that has the highest weights ...
[12:12] <gkellogg> ... Evaluating evidence that makes them think some paths are more important to others.
[12:13] <gkellogg> Ashok: what they do is create a path with parts of the protein, of the molecule, and then find a path?
[12:13] <zwu2> q+ to ask how they assign weights...
[12:13] <gkellogg> Patrick: No, what they have is interactions they know happen, and the paths are what they think are the reasons they're interactive. They do chemical studies to determine if it changes the ways that the possible paths work.
[12:14] <gkellogg> ... They collect statistical data about interactions and make paths. Having done that, they need to find the one with the best score: what is likely to be the right one.
[12:14] <gkellogg> ... They start off with a large interaction network; having mapped them, they have to find the best one.
[12:15] <gkellogg> Zwu2: I'm wondering how they assign weights among proteins and compounds.
[12:15] <gkellogg> Patrick: I don't recall the specifics of that.
[12:15] <gkellogg> Zwu2: presumably they assign weights and find a shortest-path algorithm to find the most likely path.
[12:16] <gkellogg> Ashok: did you read the paper called something like "Graph Analytics in Bio-Metrics?"
[12:16] <gkellogg> ... That's a paper that talks about a number of different path-based algorithms in biology.
[12:16] <gkellogg> ... I'll find it and send it out.
[12:17] <gkellogg> ... Let's talk briefly about Michael's case. Did you guys happen to take a look at it?
[12:17] <gkellogg> .. I sent this out as use-case 3.
[12:17] <gkellogg> Zwu2: I briefly looked at it.
[12:18] <gkellogg> ... I think it's a pretty good use case; It's one of the most important use cases for e-commerce sites.
[12:18] <gkellogg> ... If you're browsing, you may get an email or an ad based on what you are looking at and past interactions.
[12:19] <gkellogg> ... It's basically shortest-path with a lot of constraints. For example, avoiding a highway toll, or not go by sea/air.
[12:20] <gkellogg> Ashok: I thought the interesting case was if you're FedEx, and you have 100 packages to deliver and you do a couple of things. You need to cluster them, then you need to find the optimal route through them (avoid tolls, or whatever).
[12:20] <gkellogg> ... I thought that was a good one.
[12:21] <gkellogg> ... I'm going to take these use cases (3 of them), but I think Michael's may be a couple of use cases, and put them in the wiki where we can add or annotate them.
[12:21] <gkellogg> Topic: Organizing Use Cases
[12:22] <gkellogg> Ashok: if you look at the graph-theory literature, there are about 8 algorithms.
[12:22] <zwu2> This may be useful: http://bioinformatics.oxfordjournals.org/content/28/12/i49.full.pdf+html
[12:22] <gkellogg> ... clustering, dispersion, shortest-path, page-rank, traveling-salesman.
[12:22] <zwu2> "A single source k-shortest paths algorithm to infer regulatory pathways in a gene network"
[12:23] <gkellogg> Ashok: we could talk with companies which implement them, and consider the use cases for each of these algorithms.
[12:23] <gkellogg> ... If anyone has use cases for the specific algorithms, that would be great.
[12:24] <gkellogg> Zwu2: I talked with some folks from Thompson-Reuters; they also have some challenging graph problems.
[12:24] <gkellogg> ... They're dealing with cutting-edge problems.
[12:24] <gkellogg> ... Even in publishing, articles reference each other and form a graph; how do you analyze this graph?
[12:24] <gkellogg> Ashok: for what purpose?
[12:25] <gkellogg> Zwu2: recommendation, clustering, and to find out new trends.
[12:25] <gkellogg> ... I'll ping the team and see if they'd like to join the group.
[12:25] <gkellogg> Ashok: if they could just tell us how they are using using Property Graphs, that would be great.
[12:26] <gkellogg> ... Patrick has been speaking to people in biology area to get them to join too.
[12:26] <gkellogg> ... Kelvin, could you write up the use case?
[12:26] <gkellogg> Kelvin: I was going to suggest use cases for "best-performer" and so forth.
[12:27] <gkellogg> Ashok: I'll stick one in the wiki and we can add others.
[12:27] <gkellogg> Ashok: okay, we'll close the call, and speak again on 3 December.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[12:28] <Ashok> Wiki is at:http://www.w3.org/community/propertygraphs/wiki/Main_Page <http://www.w3.org/community/propertygraphs/wiki/Main_Pagehttp://www.w3.org/community/propertygraphs/wiki/Main_Page>
[12:29] <gkellogg> ... When we get Ivan again, I'd like to ask questions like how much evidence do we need to collect, etc.?
[12:29] <zwu2> thanks & bye
[12:29] <Ashok> ADJOURNED
-- 
All the best, Ashok