If I understand correctly, this algorithm tests an element of a set not just with other elements but with all subsets of elements in the set which it isn't in? I understand the reasoning, but the the number of tests for a set of N elements is 2^(N-1)-1. Yikes.
I try to come up with the matches that maximises. While for a small number of orders i would look at every permuatation practically speaking i cannot so i need to attempt it using things like genetic algoritmns and also some heuristics to reduce the size of the search space (eg consolidate this order with this order as they partially match). Also from a procedural point of o only view i was originally constrained by the fact that i had just got a fixed amount of time each week to perform the matching. However as a lot of orders are there for quite a while (meaning what is there on a monday is likley to be there on a friday and also because new orders entered during the week are very likley to be there at the end of the week then what i can do is i can be doing calcuations for the matches during the week and taking into account incremental information as it arrives. The risk of course with this is that you have a very long chain of matches and somebody cancels an order and whoosh the whole chain is invalid. One way to get around it would be to only build such chains if you had an alternate order at each point. However this would probably only add a while new set of problems. Anyway back to the project i am mostly interested in people who could do the front end. I have done the majority of the work on the backend and by all means people are free to improve on that but at this stage the crirical point is the front end. Regarding Amazon the ideal thing would be to have the website with Amazon Associate links as it would then generate revenue for the site. tks, PJ On Fri, Feb 13, 2009 at 1:26 PM, Steve McConville < [email protected]> wrote: > > 2009/2/13 PJ Fitzpatrick <[email protected]>: > > Hi, > > > > You mention a very critical point about the network. I worked on the > launch > > of an online exchange many years back and a lot of the success has to do > > with how many users you get. However i'm also of the opinion that the > actual > > matching algorithmn can have a critical bearing on the success as well. > > I'm closer to the data-centric viewpoint myself, as expressed by Anand > Rajaraman > > http://anand.typepad.com/datawocky/2008/03/more-data-usual.html > > > For example assume you can only match with someone who is the exact > opposite > > to yourself. I have A and want B and you have B and want A. To find > matches > > you then need a lot of users. If however you widen it out to allow say 3 > > people to match then you could have a match created if the following > occurs. > > -I have A and want B > > -Another user has B and wants C > > -Yet another user has C and wants A > > > > So by allowing these higher order matches you get a much better chance of > > matches actually happening for a smaller number of participants. In my > local > > barter website i have taken this to the extreme that i have no limit on > the > > number of people that can be matched in such a chain. All i do is, at > some > > fixed interval(currently expected to be weekly), i take all the orders > and > > try to come up with the matches that maximises (or attempts to maximise > as > > it is unlikley i would get the global maximum) the amount of business > done > > on the exchange. Of course things becomes impractical if the number of > users > > is very high but as i said it is a local barter website. > > If I understand correctly, this algorithm tests an element of a set > not just with other elements but with all subsets of elements in the > set which it isn't in? I understand the reasoning, but the the number > of tests for a set of N elements is 2^(N-1)-1. Yikes. > > > Anyway one of the reasons i posted this in the first place was to see > what > > people though and to see is some other people would be interested in > > participating in an Open Source Project on it. I remember a while back > there > > was talk of some project in the Python Ireland group. > > I would certainly be interested in helping to develop a project aimed > at creating a knowledge base like this, though I don't know how much > effort I could commit to. Of course the first problem such a site > would face is deciding which stack to use :P > > > Also a similar concept could be applied to books. I already have got the > > domain www.ihavereadthebook.com. > > Amazon allow third party use of their recommendation engine. You might > want to have a look at that if/when your book project kicks off. > > -- > steev > http://www.daikaiju.org.uk/~steve/ <http://www.daikaiju.org.uk/%7Esteve/> > > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Python Ireland" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.ie/group/pythonireland?hl=en -~----------~----~----~----~------~----~------~--~---
