Data Depth: Methodology and Computation

Add to your list(s) Download to your calendar using vCal

• Pavlo Mozharovskyi (Telecom Paris, Institut Polytechnique de Paris)
• Tuesday 06 May 2025, 10:15-11:15
• Seminar Room 1, Newton Institute.

If you have a question about this talk, please contact nobody.

RCLW01 - Uncertainty in multivariate, non-Euclidean, and functional spaces: theory and practice

Data depth is a statistical function that measures centrality of an observation with respect to a probability distribution, with an empirical measure (on a data set) being its most important particular case. By exploiting the geometry of data, the depth function is fully non-parametric, robust to both outliers and heavy tailed distributions, and satisfies desirable invariances. By dint of these advantages, it is used in a variety of tasks as a generalisation of quantiles in higher dimensions and as an alternative to the probability density. Introduced in the second half of the twentieth century and having undergone theoretical and computational developments since then, data depth became a universal methodology for ordering complex data and is now employed in numerous applications: supervised and unsupervised machine learning, robust optimisation, financial risk assessment, statistical quality control, extreme value theory, imputation of missing data, etc. In this presentation, we should survey the notion of data depth and highlight its most relevant advantages. We shall start with the formal definition of statistical data depth function and study of its (most important) properties. This will be followed by an assortment and analysis of commonly used depth notions. Furthermore, relevant computational aspects will be regarded including most recent advances that allow depth-based applications on contemporary scale. The presentation shall be accompanied by application examples on synthetic and real data sets. Finally, several open questions will be discussed in the outlook.

This talk is part of the Isaac Newton Institute Seminar Series series.

This talk is included in these lists:

• All CMS events
• Featured lists
• INI info aggregator
• Isaac Newton Institute Seminar Series
• School of Physical Sciences
• Seminar Room 1, Newton Institute
• bld31

Note that ex-directory lists are not shown.