Abstract: Partition clustering is sensitive to initialisation. Moreover, it is not always clear whether a unique best solution exists for a given practical problem. A methodology is presented for mapping the space of partition clusters, from which to sample reproducible cluster solutions with good separation measured by second-order statistics and with stability to repeated initialisations. The cluster solutions are reproducible in the sense that multiple independent re-samplings of the initial conditions will return substantially identical partitions, for the same data and clustering algorithm. An extension of the methodology is proposed to reproducibly define a hierarchical tree of cluster solutions for increasing cluster number. The method is evaluated with synthetic data and applied to a previously studied protein expression dataset.
Page 13 of 15
Dr. Terence A. Etchells gave on April 15th 2010 a conference at IDAL facilities titled “Rules from Data via a Neural Network. Orthogonal Search-based Rule Extraction (OSRE)“. Prof. Etchells was visiting IDAL for 5 days, although visit was unintentionally extended due to air space closed because of volcanic ash. IDAL members thank professor Etchells his excelent dissertation.
Ian H. Jarman and Terence A. Etchells, post-doctoral researchers of the Liverpool John Moores University (United Kingdom) will start a research stay with IDAL in the next weeks. The main topics of the joint research will include discrete clustering and graphical models, as well as some practical applications of those methods. Dr. Jarman worked on an integrated framework for risk profiling of breast cancer patients in his Ph.D, while Dr. Etchells worked on an efficient method for rule extraction from neural networks. Currently, both researchers are especially involved with the Public Health Research Centre of the LJMU, although their research also deals with theoretical aspects of Machine Learning algorithms and their applications beyond Public Health. IDAL has collaborated with the group of Statistics and Neural Computing of the LJMU, led by Prof. Paulo J. G. Lisboa since 2003.
IDAL has recently attended a 5-day Workshop in Graphical Models, conducted by José Antonio Gámez, from University of Castilla-La Mancha, (Albacete, Spain).
José A. Gámez was born in Granada (Spain) in 1968. He received his B.Sc. (1989) and M.Sc. (1991) degree in Computer Science from the University of Granada. He got his PhD. in Computer Science in September 1998 at the Dept. of Computer Science and Artificial Intelligence (University of Granada). The topic of the thesis was Partial Abductive Inference in Bayesian Networks.
Since November 1991 he is a member of the Computer Systems Department (formerly Computer Sience Department) at the University of Castilla-La Mancha. Jose A. Gámez started as Assistant Professor and in July 1994 he got a permanent position as Lecturer. In December 2002 he became Associate Professor. During the period September-1998 to September-2004 he served as Vice-Dean in the Polytechnic School (EPSA).
His current research areas are in the field of Bayesian Networks, evolutionary computation and data mining. More info in the Intelligent Systems and Data Mining webpage.
The 2010 IEEE World Congress on Computational Intelligence (IEEE WCCI 2010) is the largest technical event in the field of computational intelligence. It will host three conferences: the 2010 International Joint Conference on Neural Networks (IJCNN 2010), the 2010 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE 2010), and the 2010 IEEE Congress on Evolutionary Computation (IEEE CEC 2010). IEEE WCCI 2010 will be held in Barcelona, a Mediterranean city located in a privileged position on the northeastern coast of Spain.
IDAL collaborates in organizing a special session in Computational Intelligence for Bioinformatics (See the complete list of special sessions here). IDAL session is in number 47.
Hope to see you soon in Barcelona!
DAL Member J. Gómez has defended today his Doctoral Thesis “Detección automática de podredumbres en cítricos mediante procesado avanzado de imágenes hiperespectrales” (Automatic detection of rotten areas in oranges by advanced hyper spectral image processing) with the maximun mark.
Comgratulations for the new doctor at IDAL!
IDAL member José D. Martin collaborate in organizing a special session in European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (ESANN), that will be held in Bruges (Belgium), 28-30 April 2010.
The session is coorganizated with Prof. Paulo Lisboa and Prof. Alfredo Vellido.
READ MORE: IDAL IN CO-ORGANIZATION OF A SPECIAL SESSION IN ESSAN
IDAL member JDMartin will attend the XXII Congress of the International Society of Biomechanics, with a work titled “Modeling dorsal pressures on the foot surface using artificial neural networks”. The congress will be held in Cape Town, South Africa from 5th to 9th July 2009. You can gather more information in http://www.isb2009.org/
IDAL member Jose D. Martín gave a invited conference on a methematical model for iron dosage in kidney failure diseases. The conference was hold at the WCN 2009 in Milan. The conference and slides can be seen online in.
http://tv.online-congress.com/fmcrenalpharma/090524fmcrenalpharma_wcn2009/martin.html
IDAL members J. Gómez and R. Magdalena hae atended the International Joint Conference on Nural Networks, hold in Atanta USA 13-19 June 09. They have presented an invited poster in a special session in CI in Direct Support of Patient Care, titledSurvival Prediction in Patients Undergoing Ischemic Cardiopathy. Charimans were Adam Gaweda, James diLeo and Alexandru Floares.