Xavier Llorà » Estimation of Distribution Algorithms

Scaling eCGA Model Building via Data-Intensive Computing

Xavier — Thu, 08 Apr 2010 16:17:39 +0000

I just uploaded the technical report of the paper we put together for CEC 2010 on how we can scale up eCGA using a MapReduce approach. The paper, besides exploring the Hadoop implementation, it also presents some very compelling results obtained with MongoDB (a document based store able to perform parallel MapReduce tasks via sharding). The paper is available as PDF and PS.

Abstract:
This paper shows how the extended compact genetic algorithm can be scaled using data-intensive computing techniques such as MapReduce. Two different frameworks (Hadoop and MongoDB) are used to deploy MapReduce implementations of the compact and extended com- pact genetic algorithms. Results show that both are good choices to deal with large-scale problems as they can scale with the number of commodity machines, as opposed to previous ef- forts with other techniques that either required specialized high-performance hardware or shared memory environments.

GECCO 2010 Submission Deadline (Extended)

Xavier — Sat, 19 Dec 2009 12:11:34 +0000

If you are planning to submit a paper for the 2010 Genetic and Evolutionary Computation Conference, the deadline is January 13, 2010 (and now extended to January 27th). You can find more information at the GECCO 2010 calendar site. Related posts: GECCO 2009 paper submission deadline extended till January 28 GECCO 2011 Submission Deadline: January [...] Related posts:

Scaling Genetic Algorithms using MapReduce

Xavier — Fri, 09 Oct 2009 15:51:19 +0000

Below you may find the abstract to and the link to the technical report of the paper entitled “Scaling Genetic Algorithms using MapReduce” that will be presented at the Ninth International Conference on Intelligent Systems Design and Applications (ISDA) 2009 by Verma, A., Llorà, X., Campbell, R.H., Goldberg, D.E. next month.

Abstract:Genetic algorithms(GAs) are increasingly being applied to large scale problems. The traditional MPI-based parallel GAs do not scale very well. MapReduce is a powerful abstraction developed by Google for making scalable and fault tolerant applications. In this paper, we mould genetic algorithms into the the MapReduce model. We describe the algorithm design and implementation of GAs on Hadoop, the open source implementation of MapReduce. Our experiments demonstrate the convergence and scalability upto 105 variable problems. Adding more resources would enable us to solve even larger problems without any changes in the algorithms and implementation.

The draft of the paper can be downloaded as IlliGAL TR. No. 2009007. For more information see the IlliGAL technical reports web site.

Data-Intensive Computing for Competent Genetic Algorithms: A Pilot Study using Meandre

Xavier — Tue, 14 Jul 2009 04:15:51 +0000

Below you may find the slides I used during GECCO 2009 to present the paper titled “Data-Intensive Computing for Competent Genetic Algorithms: A Pilot Study using Meandre”. An early preprint in form of technical report can be found as an IlliGAL TR No. 2009001 or the full paper at the ACM digital library

GECCO 2007 Hat Trick

Xavier — Fri, 13 Jul 2007 03:46:07 +0000

Yup, never happened to me before. In John Koza’s words: Hat Trick, best paper in EDA track and two Bronze medals in the HUMIES competition. I just think that I was lucky and blessed to have the pleasure to work with such great people. It would not have happened if it wasn’t for them. Below you can find some of the presentations for the above Hat Trick, as well as some other work I presented on genetics-based machine learning.

Evolving emotional prosody

Xavier — Sun, 17 Sep 2006 15:57:30 +0000

by Cecilia Ovesdotter Alm and Xavier Llorà (2006).
Proceedings of the Ninth International Conference on Spoken Language Processing (INTERSPEECH 2006), paper 1741. Also as IlliGAL TR No 2006018. Link to the PDF.

Abstract
Emotion is expressed by prosodic cues, and this study uses the active interactive Genetic Algorithm to search a wide space for sad and angry parameters of intensity, F0, and duration in perceptual resynthesis experiments with users. This method avoids large recorded databases and is flexible for exploring prosodic emotion parameters. Solutions from multiple runs are analyzed graphically and statistically. Average results indicate parameter evolution by emotion, and appear best forsad speech. Solutions are quite successfully classified by CART, with duration as main predictor.

E2K: Evolution to knowledge

Xavier — Wed, 13 Sep 2006 20:14:15 +0000

by Xavier Llorà (2006).
ACM SIGEvolution, Volume 1 , Issue 3, pp. 10-17. Link to the Journal.
Also as IlliGAL TR No 2006022. Link to the PDF.

Abstract
Evolution to Knowledge (E2K) is a set of Data to Knowledge (D2K) modules and itineraries that perform genetic algorithms (GA) and genetics-based machine learning (GBML) related tasks. The goal of E2K is to fold: simplify the process of building GA/GBML related tasks, and provide a simple exploratory workbench for the evolutionary computation community to help users to interact with evolutionary processes. It can help to create complex tasks or help the newcomer to get familiarized and trained with the evolutionary methods and techniques provided. Moreover, due to its integration into D2K, the creation of combined data mining and evolutionary task can be effortlessly done via the visual programming paradigm provided by the workflow environment and also wrap other evolutionary computation software.

GECCO 2006

Xavier — Sat, 08 Jul 2006 18:14:50 +0000

Well, it finally happened. GECCO started today at 8:30am. You can find live blogging at the IlliGAL Blogging. Related posts: GECCO 2007 Hat Trick Yes, GECCO 2008 was intense! GECCO 2010 Submission Deadline (Extended) Related posts:

The χ-ary extended compact classifier system: Linkage learning in Pittsburgh LCS

Xavier — Fri, 07 Jul 2006 06:00:38 +0000

by Xavier Llorà, Kumara Sastry, David E. Goldberg, and Luis de la Ossa (2006).
To appear in the Proceedings of the International Workshop on Learning Classifier Systems (IWLCS 2006). Also as IlliGAL TR No 2006015. Link to the PDF.

Abstract
This paper proposes a competent Pittsburgh LCS that automatically mines important substructures of the underlying problems and takes problems that were intractable with first-generation Pittsburgh LCS and renders them tractable. Specifically, we propose a χ-ary extended compact classifier system (χeCCS) which uses (1) a competent genetic algorithm (GA) in the form of χ-ary extended compact genetic algorithm, and (2) a niching method in the form restricted tournament replacement, to evolve a set of maximally accurate and maximally general rules. The results clearly show that linkage exists in the multiplexer problem which needs to be accurately discovered and efficiently processed in order to solve the problem in tractable time. The results also show that in accordance with the facetwise models from GA theory, the number of function evaluations required by χeCCs to successfully evolve an optimal rule set scales exponentially with the number of address bits (building block size) and quadratically with the problem size.

A simple UMDAc implementation in Java

Xavier — Tue, 06 Dec 2005 06:04:55 +0000

Cecilia Oversdotter is working on an adaptation of active interactive genetic algorithms (some papers can be found here and here) to her work on speech synthesis and perception of emotions in expressive storytelling. She needs a version of the active interactive genetic algorithm that works on continuous domains. For that reason I coded a version of UMDAc to replace the cGA currently used for discrete domains.

The Java implementation of UMDAc can be found here. In order to run it, you need to download the COLT toolkit . The code is distributed under GPL license.