DECIE2011: Distributed Evolutionary Computation using REST

This paper analises distributed evolutionary computation based on the Representational State Transfer (REST) protocol, which overlays a farming model on evolutionary computation.
An approach to evolutionary distributed optimisation of multilayer perceptrons (MLP) using REST and language Perl has been done. In these experiments, a master-slave based evolutionary algorithm (EA) has been implemented, where slave processes evaluate the costly fitness function (training a MLP to solve a classification problem).
Obtained results show that the parallel version of the developed programs obtains similar or better results using much less time than the sequential version, obtaining a good speedup.

Advertisements

DECIE2011: SOAP vs REST: Comparing a master-slave GA implementation

In this paper, a high-level comparison of both SOAP (Simple Object Access Protocol) and REST (Representational State Transfer) is made. These are the two main approaches for interfacing to the web with web services.
Both approaches are different and present some advantages and disadvantages for interfacing to web services: SOAP is conceptually more difficult (has a steeper learning curve) and more ”heavy-weight” than REST, although it lacks of standards support for security.
In order to test their eficiency (in time), two experiments have been performed using both technologies:
a client-server model implementation and a master-slave based genetic algorithm (GA).
The results obtained show clear differences in time between SOAP and REST implementations.
Although both techniques are suitable for developing parallel systems, SOAP is heavier than REST, mainly due to the verbosity of SOAP communications (XML increases the time taken to parse the messages).

The bestest MasterMind Algorithm ever

Well, by now, you must be a bit tired of Mastermind papers but we are not, since we are obtaining the best results ever. After introducing end games to streamline the end of the algorithms, we have tweaked the evolutionary algorithm, adding a permutation operator, for instance, to reduce the number of evaluations needed to find the solution. The results is the best yet, but, of course, there’s more to come in the future.
This paper was presented at CEC 2011 in the games session, and raised quite a bit of interest. The paper will be available from IEEExplore soon, but you can request copies now if you want

Doing evolutionary algorithms with Dropbox

Why not use Dropbox as its name implies, as a box for dropping individuals that could be interchanged among different islands running evolutionary algorithms?
That’s exactly what we are doing in a series of papers that are being published and presented in IWDECIE, CEC 2011 and GECCO, in last-in, first-out order. This presentation is for the second, presented today in CEC.

What we try to test in this paper is whether we can add a good number of computers (up to 4) without a saturation of the network (or of Dropbox itself), and whether there is a difference between wired and wireless. It so happens there is, but it gets smaller when you increase the number of computers.
IMG_0189
Still many tests to to, but for the time being this looks promising. We’ll link the paper when it’s available. For the time being, if you’re interested just send us an email