TSP is generally solved using a number of heuristics guiding a randomized search. Mahout has essentially no provision for helping with this.
If you want a quick and dirty solution, I would recommend something like an evolutionary algorithm in which you have segments that self-assemble or split with the parameters controlling the assembly and splitting subject to auto-mutation. Conventional genetic algorithms should also work reasonably well, but you will almost certainly have to include some auto-evolution. Mahout does have a directed-step evolutionary algorithm, but it is not suitable for discrete problems such as TSP. On Fri, Jan 10, 2014 at 1:11 AM, Pavan K Narayanan < pavan.naraya...@gmail.com> wrote: > Hi > > Has anyone tried solving travelling salesman problem using Mahout? > (could be any version) . If yes, I would like to know, > > 1. What is the format of input data? is it txt file or csv file or any > other format. > 2. do you have to divide the input data into so chunks? for example, > if your tsp has 10k cities and you divided that into two 5k cities > problems. (I understand running on hadoop will divide the data into > chunks of 64mb or any other user defined, but was there any external > division. > > I have not solved TSP using Mahout. Would appreciate if anyone could > walk me thro the process of solving tsp. > > thanks >
