Projection pursuit is a multivariate statistical technique aimed at detecting interesting low-dimensional data projections. It looks for the data projection which maximizes the projection pursuit index, that is a measure of its interestingness. After an interesting projection is found, it is removed to facilitate the search for other interesting features. Projection pursuit addresses three major challenges of multivariate analysis: the curse of dimensionality, the presence of irrelevant features and the limitations of visual perception. Its applications have been hampered by computational, interpretative and inferential problems. Additional problems arise when data are high-dimensional, that is when there are more variables than units. This session outlines the main features of projection pursuit and its connections with other multivariate techniques. The theory is illustrated with both real and simulated datasets.
Organized by:
Nicola Loperfido (Italy)A moment-based projection pursuit index
Invited Speakers:
- Andriette Bekker (South Africa)
Modeling incomplete compositional datasets - Alessandro Berti (Italy)
Projection Pursuit for Detecting Hidden Structures - Claudio Borroni (Italy)
On the use of some unconventional projection pursuit indexes in cluster identification problems - Manuela Cazzaro (Italy)
Competing projections techniques of multivariate quality data to monitor a production process - Lucio De Capitani (Italy)
On copula-based systems of quantile curves with application in ouliers detection - Cinzia Franceschini (Italy)
Some remarks on kurtosis projection pursuit for clustering - Nicola Loperfido (Italy)
A moment-based projection pursuit index - Marco Morosin (The Netherlands)
Modeling high-dimensional data with a multivariate Bernoulli distribution - Boaz Nadler (Israel)
A simple method for robust matrix completion with theoretical recovery guarantees - Jaakko Pere (Finland)
On Stationary Subspace Analysis for Spatio-Temporal Data - Perttu Saarela (Finland)
Stationary subspace analysis for spatial data - Tomer Shushi (Israel)
Optimal portfolio projections: Optimizing portfolio returns in the case of elliptically and skew-elliptically distributed models