Machine Learning

LCSweb was designed to allow researchers and those seeking to use Learning Classifier Systems within applications access to material on LCS and discussion between members of the LCS community. The site served this community since its was started by Alwyn Barry in 1997. Enhanced and maintained later by Jan Drugowitsch, LCSweb became a valuable community resource. The site was completely community-driven and allowed members to contribute to the content of the site and keeping it up to date. Later on in 2005, I started “LCS and other GBML” Blog to cover a gap providing information information regarding the International Workshop on Learning Classifier Systems (IWLCS), the collection of LCS Books available, and GBML related news. Some of you may have realized that after Jan’s move to Rochester and Alwyn’s retirement from research activities, LCSweb has vanished. Will Browne took on himself to take LCSweb to Reading, but technical circumstances have made that move rocky despite his best efforts. Jan and Will however still have a local copy of LCSweb contents. After talking to Jan and Will, I proposed to merge LCSweb with the LCS and other GBML blog, and host the new site at NCSA where dedicated resources has been made available. Jan and Will agreed with the idea. We are happy to announce that the merged site (still under the update cycle) can be reached at http://lcs-gbml.ncsa.uiuc.edu. More information about the process can be found here or at there LCS & GBML Central site. ...

After attending Albert Orriols’s Ph.D. thesis defense, I ended wondering how many of the question I posted in mine have not been solved. The answer, quite a bit. So, I just decided to dig it up, and put it up here. Yes, the thesis was not written in English (in those days my fellowship had some strings attached), but math formulation, graphs, and results are readable in any language ;) Also, GALE was written and documented in english, and is available here. ...

Albert Orriols (for those who do not know him, a brilliant learning classifier systems researcher) defended his thesis today. The outcome: Excellent Cum Laude. Albert Orriols Ph.D. defense started at 11am at Enginyeria i Arquitectura La Salle in Barcelona. The thesis panel was presided by Prof. David E. Goldberg, and formed by members Prof. Francisco Herrera, myself, Dr. Martin Butz, and secretary Prof. Xavier Vilasís). I must say that it has been a great pleasure to read his remarkable thesis and great contributions to the Learning Classifier System field. I hope he will make it available soon, and encourage you to take a look at it. ...

by Xavier Llorà, Anusha Priya, and Rohit Bhargava (2006). To appear in the Special Issue on Learning Classifier Systems of the Natural Computing Journal. Also as IlliGAL TR No. 2006027. Link to the PDF. Abstract Prostate cancer accounts for one-third of noncutaneous cancers diagnosed in US men, and it is a leading cause of cancer-related death. Advances in Fourier transform infrared spectroscopy of stained tissue is now able to provide very large data sets describing the chemical properties of the cells forming the prostate tissue. Uniting spectroscopic imaging data and computer-aided diagnoses (CADx), we seek to provide a new approach to pathology by automating the recognition of cancer in complex tissue. The first step toward the creation of such CADx tools requires mechanisms for automatically learn tissue type classification—a key step on the diagnosis process. As we will show, genetics-based machine learning (GBML) can be used to approach such a problem. However, there is an urge for efficient and scalable implementations that enable to process such very large data sets. This paper proposes and validates and efficient GBML technique—NAX—based on an incremental genetics-based rule learner that exploits massive parallelisms—via the message passing interface (MPI)—and efficient rule-matching using hardware-implemented operations. Results show the competence of NAX solving the prostate tissue type prediction and how such and efficient implementation makes it a very powerful tool for biomedical image processing. ...

by Xavier Llorà and Kumara Sastry (2006, accepted). Proceedings of the ACM Genetic and Evolutionary Computation Conference (GECCO 2006), pp. 1513–1520, ACM press. Also as IlliGAL TR No 2006001. Link to the PDF. Abstract Over the last ten years XCS has become a de facto standard for Michigan-style learning classifier systems (LCS). Since the initial CS-1 work conceived by Holland, classifiers (rules) have widely used a ternary condition alphabet {0,1,#} for binary input problems. Most of the freely available implementations of this ternary alphabet in XCS rely on character-based encodings—easy to implement, not memory efficient, and expensive to compute. Profiling of freely available XCS implementations shows that most of their execution time is spent determining whether a rule is match or not, posing a serious thread to XCS scalability. In the last decade, multimedia and scientific applications have pushed CPU manufactures to include native support for vector instruction sets. This paper presents how to implement efficient condition encoding and fast rule matching strategies using vector instructions. The paper elaborates on Altivec (PowerPC G4, G5) and SSE2 (Intel P4/Xeon and AMD Opteron) instruction sets producing speedups of XCS matching process beyond ninety times. Moreover, such a vectorized matching code will allow to easily scale beyond tens of thousands of conditions in a reasonable time. The proposed fast matching scheme also fits in any other LCS other than XCS. ...