[Solomonov Seminar] 171. Solomonov seminar
marko.grobelnik at ijs.si
Mon Jun 19 11:08:57 CEST 2006
Vabim vas na 170. Solomonov seminar, ki bo jutri, v torek 20. maja,
ob 13:00 uri v Veliki predavalnici IJS.
Predaval bo nas gost iz Trenta Giuseppe Jurman na temo uporabe
strojnega ucenja v bioinformatiki in genomiki.
"Machine Learning for High-Throughput Functional Genomics"
Giuseppe Jurman (ITC - irst, Trento, Italy http://mpa.itc.it/jurman/)
Predictive classification and feature selection and ranking
are two main tasks when looking for molecular profiles.
Although powerful machine learning tools are available to deal
with highly dimensional data, carefully designed experimental setups
are essential to avoid overfitting effects (such as the selection bias).
Complex validation procedures and computing resources are thus required
in order to obtain honest estimates of accuracy on novel data and
correct selection of biomarkers.
In this talk I will outline the methodology and the computational solutions
implemented in the BioDCV system, a complete tool for high-throughput
data analysis, together with a few examples of recent tasks on microarray
and proteomics data. Finally, other key issues such as biomarker selection
and stability, discovery of outliers and of potential subtypes
and treatment of time-varying patterns will also be discussed.
(Joint work with C. Furlanello, S. Merler, A. Barla, S. Paoli, D. Albanese, B. Irler, R. Flor)
More information about the Solomonov-seminar