HDR UK Applied Analytics Webinar: Relevance, redundancy and complementarity trade-off (RRCT) - Thanasis Tsanas (45:15)

The presence of a large number of features in a dataset is often detrimental in the statistical learning process. This is widely known as the curse of dimensionality and occurs because the feature space is inadequately populated and/or the decrease of the inherent signal to noise ratio in the dataset. Many feature selection techniques have been developed over the years aiming to identify a robust feature subset aiming to develop a parsimonious model. In this talk, I will describe my latest work on developing an information theoretic algorithm and demonstrate how it performs competing against 20 feature algorithms across 12 synthetic and real-world datasets. MATLAB and Python source is available in my research group’s website: https://www.darth-group.com; I will demonstrate how this can be run off-the-shelf and how it can be easily deployed in different applications. (Dr Thanasis Tsanas) 30/06/22