Machine Learning

SVM + Patch-Based Features

Saadia Iftikhar worked on approaches to image segmentation which rely on specifying a rich set of visual features, feature selection and machine learning. The specification of user-labelled examples is required, but this makes the approach suitable for a wide range of biological segmentation problems, allowing expert users to train a system which, based on the non-linear SVM, generalises well within that image class. SVMs are able to define highly non-linear boundaries in feature space.   More recently,t he technique was applied to segmentation of retinal vessels as well by Shearin Cao.  See the Publications page for more details.

Tutorial: A Shallow Introduction to Deep Learning

This is based on a tutorial given at the NETT Induction Workshop held in Barcelona in November,  2013, an event within a Marie-Curie ITN. See here for the updated  presentation (PDF format).  You can also find a bibliography on Deep Learning, collated by Woo-Sup Han [HTML+BibTeX].  There is also a PDF version, with a short forward, at the link below: [wpdm_package id=’1279′]