Jorge Tavares weblog

Posts Tagged ‘Evo*

Best Paper Award at EuroGP 2012

with 4 comments

Last week I was at EuroGP where I presented my latest paper “Automatic Design of Ant Algorithms with Grammatical Evolution” (pdf, slides) and it won the Best Paper Award! I am very happy with this distinction since EuroGP is the leading conference on Genetic Programming attended by the best researchers in the field.

In this paper, a Grammatical Evolution approach is used for the automatic design of Ant Colony Optimization algorithms. The used grammar has the ability to guide the learning of novel architectures, by rearranging components regularly found on manually designed variants (for example, the Elitist Ant-System, the Ant Colony System or the Max-Min Ant System). This approach was tested with the TSP and the results show that the evolved algorithmic strategies are effective, exhibit a good generalization capability and are competitive with human designed variants. This is still a starting point and there is a large amount of work to be done but the indications given by these results are encouraging!

This year the conference took place in Málaga, Spain, as part of Evo* as usual. The event was great with an excellent organization lead by Carlos Cotta. It was a fantastic Evo*! Next year it will be held in Wien!

Written by Jorge Tavares

April 17, 2012 at 20:51

Evo* 2011: brief notes

leave a comment »

Last week I was in Turin for Evo* 2011. It was a nice event and it is always good to meet again friends and collegues. However, Evo* has lots of parallel sessions and so it becomes impossible to see everything you want or would like. This year I decided to focus a bit more than previous times and just attended EuroGP sessions (several), EvoCOP (two) and EvoSTOC (one). Although some of the talks were interesting I didn’t see anything really out of extraordinary. Still, you always learn something new. The invited talks were also ok but lacked that inspiring touch, as oposed to the ones in PPSN last year. Regarding my talk, the feedback was positive and I had two ideas that might be good to experiment with. As for the organization, it was the usual but the lunches were really not that great (my only complain I must say). Next year it’s going to be in Málaga, Spain, organized by Carlos Cotta and J.J. Merelo. I think we can only expect a great event!

Written by Jorge Tavares

May 2, 2011 at 9:03

Posted in Research

Tagged with ,

EuroGP 2011 Best Paper Nomination

with 4 comments

Just received an email with good news! The paper that I got accepted at EuroGP was nominated for Best Paper! The work that I started last year is starting to get some good feedback which is great :-) This paper describes the approach I have developed (in collaboration with Francisco B. Pereira) to evolve pheromone update methods for Ant Systems. The paper only deals with the TSP and the obtained results are promising since it outperforms MMAS. It is the first step to what I am working on, a Self-Ant System. The idea is to let the Ant algorithm find the best components that can help in solving the problem at hand. There is still a lot of work to be done but this is just the start.

Written by Jorge Tavares

March 3, 2011 at 15:26

Going to Evo* in Istanbul

leave a comment »

Although a bit late for this post but the year started with a good news: the paper I co-authored with Salma Mesmoudi was accepted at EvoBIO 2010. So, I am going to Istanbul to attend the Evo* set of conferences and workshops. This paper results from my final time at INRIA with Salma. We are still collaborating but due to my constant movements we weren’t able to finish it sooner. Let’s see if we can continue to work on some of possibilities that are open with this work.

And for the record, our paper is titled “Variable Genetic Operator Search for the Molecular Docking Problem”, and the abstract is:

The aim of this work is to present a new hybrid algorithm for the Molecular Docking problem: Variable Genetic Operator Search (VGOS). The proposed method combines an Evolutionary Algorithm with Variable Neighborhood Search. Experimental results show that the algorithm is able to achieve good results, in terms of energy optimization and RMSD values for several molecules when compared with previous approaches. In addition, when hybridized with the L-BFGS local search method it attains very competitive results.

[UPDATE]: Just received the news that the paper was nominated for the Best Paper Award!

Written by Jorge Tavares

February 23, 2010 at 8:32

Follow

Get every new post delivered to your Inbox.