Berthold Lausen has recently led a seminar titled ‘Ensemble of selected classifiers’ in Norwich.
We are delighted that one of our researchers, Berthold Lausen, was asked to present a seminar at the Sainsbury Centre for Visual Arts in Norwich. He discussed his recent proposals to improve classification based on high dimensional feature space. For example after preprocessing microarray data with 500,000 probes and 22,125 features (probesets) which represent genes, he used his proposal to improve feature selection of microarray data based on a proportional overlapping score (Mahmoud et al. 2014). By using benchmark data sets he will compare random forests and his recent proposals of new classification methods based on ensembles of selected k-nearest neighbours and tree classifiers (Khan et al. 2016; Gul et al. 2016).
Gul, A., Perperoglou, A., Khan, Z., Mahmoud, O., Miftahuddin, M., Adler, W., Lausen, B. (2016), Ensemble of a subset of kNN classifiers, Advances in Data Analysis and Classification.
W., Lausen, B. (2016), An ensemble of optimal trees for class membership probability estimation. In: Wilhelm, A., Kestler, H. (eds.), Proceedings of ECDA2014, Springer, Heidelberg.
Mahmoud, O., Harrison, A.P., Perperoglou, A., Gul, A., Khan, Z., Metodiev, M., Lausen, B. (2014), A feature selection method for classification within functional genomics experiments based on the proportional overlapping score, BMC Bioinformatics, Volume 15 (1).