Hi, I agree with you that bringing my sister to prom is far from cool, and I would rather go alone or not go to prom at all. But I would never have learned the Viennese waltz without my sister. Imagine going to the prom with your super hot robot with scuds on her chest[1] and making a complete fool of yourself for not being able to do a simple foxtrot. I'm sure your glad you at least have a sister! Also, building a robot that knows how to dance, and not only looks cool, is hard work and takes some time.
What I am saying is that I agree with you. We should definitely put an effort into creating good algorithms that people will use, and that will make their applications rock. But for a first algorithm package on top of Neo this is not bad. Far from bad. This is good. It might not be show material, and it might not be something to run live on huge graphs (maybe for offline analysis). But it is a very good start, and a good base line for comparisons once we get the mind blowing state of the art algorithms. I know this is one of your areas of expertise Marko, and all of us at Neo technology will be happy to listen to any valuable input from you. I have your publications on my to read list, and this email prompted me to bring them to the top of that list. I will read them as soon as I can after getting back to Sweden. If you could point me to further reading that you have found valuable I would be happy to hear about that. We would also be happy to collaborate with your algorithm work as much as possible. Let's build even groovier fembots together and sweep the floor at prom. Your algorithms and our database, I'm sure that would rock JavaOne. Take care, Tobias [1] http://www.tankgirl.info/tankgirl/images/strips/covers/tg/tgcov1.jpg On Sat, Sep 27, 2008 at 8:12 PM, Marko A. Rodriguez <[EMAIL PROTECTED]> wrote: > Hello Partik (everyone), > > I noticed that you published your algorithms component to Neo. While I > think your development looks nice (good organization to the API), there is > one primary concern I have -- (not that my concerns should be any of > yours). For instance, here is the list of graph analysis algorithms that > I grabbed off of your docs: > > * Eccentricity > * Network diameter > * Network radius > * Stress centrality > * Closeness centrality > * Betweenness centrality > * Eigenvector centrality > > See anything strange about all this? No? Okay, well here it is: all of > these algorithms are intractable on any decent sized graph. > > The future of graph analysis is not using outdated algorithms designed by > social scientists in 1960s to anaylze 15 node networks with pen and paper. > While these are the algorithms that the people *know*, they aren't the > algorithms that people will *use*. With Neo scaling to a billion edges, > all of these algorithms are dead in the water. Pointless. While they may > look good on an API and would work great for a <10,000 edge graph > (assuming every last clock cycle is accounted for in the code), for > someone using Neo as a graph *database* (not a graph analysis toolkit), > these classes might as well be interfaces cause whose going to run them? > > Again, I stress this point because I care. Neo is a completely different > beast than your typical JUNG, NetworkX, iGraph, etc. Its not a "toolkit" > --- its a database. And there are few such graph databases out there in > the world. To take the "toolkit" approach when designing an algorithms > package for a database just isn't being creative with the problem space. > > Here is an analogy: Its like bringing your sister (out dated graph > algorithms) to a high school dance (Java One). Yes you are too fat > (database) and no girls want to go with you. You don't have enough time to > slim down (toolkit), so your mom (Emil) forces your sister (out dated > graph algorithms) to go with you. Yes you can go with your sister, or, if > you were cool, you'd engineer some sweet robot monster with laser beams > for eyes (sick graph algorithms) and show up at the dance (Java One) > showing everyone that fat (databases) is the new slim (toolkits). > > Hope that makes it clear... > > Take care, > Marko. > _______________________________________________ > Neo mailing list > User@lists.neo4j.org > https://lists.neo4j.org/mailman/listinfo/user > -- Tobias Ivarsson <[EMAIL PROTECTED]> Hacker, Neo Technology www.neotechnology.com Cellphone: +46 706 534857 _______________________________________________ Neo mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user
