Calot.info repository

Earlier this year I’ve successfully defended my Engineering degree thesis. The making of took me about two years and produced several papers that were published as book chapters or articles.

The thesis was entitled “Pattern recognition on Medical Images based on Intelligent Systems” (in Spanish) and makes an approach to mammography anomalies detection using back-propagation neural network algorithms and other image processing filters like Sobel operator. After comparing the proposed method with former ones I found out that I had achieved 89% of success rates compared to 60% obtained by previous work. I’m still not so happy with the result, I believe a 99% may be achieved very easily.

The produced papers where

• Calot, E. et al. 2008. Clasificación de Tumores en Mamografías Mediante Uso Combinado de RBP y Filtros Sobel. Proceedings del XIV Congreso Argentino de Ciencias de la Computación, Congreso General, Artículo 1875. ISBN 978-987-24611-0-2 (in Spanish).
• Calot, E. et al. 2009. Tumor Classification on Mammographies Based on BPNN and Sobel Filter. In Opportunities and Challenges for Next-Generation Applied Intelligence. Studies in Computational Intelligence Series, Volume 214. pp. 19-24. ISBN 978-3-540-92813-3 (in English).

In both cases I was unable to travel. In the first one was in Chilecito, Argentina, but I had a schedule issue due to the presentation of the repertory grid paper. The second one was in Tainan, Taiwan but another scheduling issues and the swine flu outbreak made the trip impossible.

I’ve been very interested in Back Propagation Neural Networks. It is a very interesting kind of neural network capable of “remembering” stuff and making interesting associations between a previously-known input and output data (called training set) to use them to “predict” output results from given unknown input.

Here I’m linking a small “unpublished” article I wrote to clarify the training and running process (in Spanish). I’ll be working in an English version ASAP.

During this year I’ve been working in a paper on software acquisition for the Intelligent Systems Lab in the Engineering College at UBA.

This paper is the follow up of an old paper written before, improving the old method and obtaining wonderful results. The work was accepted by IFIP and presented at the IFIP World Computer Congress in Milan, Italy on September, the 9th. In this opportunity four lab team members -myself included- were able to travel to present the papers.

This work is a sort of draft on how the Kelly‘s repertory grid method could be improved. Optimizing relationships shows out a set of weaknesses the original method had and proposes some workarounds.

The past months I’ve been involved in the publication of a nice paper exploring how a machine can create good lies to reach it’s goals. The produced document has been published and exposed in the Ibero-American Symposium on Software Engineering. You can read the Original Paper Article.