Artificial Intelligence

Biography

Biography: Fionn Murtagh

Abstract

The book with the title, “Th e Sciences of the Artifi cial”, is by Nobel Prize winner in 1978, Simon Herbert. At issue is cognitive processes and analytics from inherent hierarchical system complexity. We may be determining the extent of hierarchical nature and properties, perhaps including or determining evolution along the lines of geneology. First we address how inherently hierarchical various sources of data can be. Considered are time series that are fi nancial, environmental, biomedical, and texts that are from literature, from accident reports, and psychogically related dream reports. Th e use and benefi t of taking hierarchical structure fully into account includes the following: how high dimensional or sparse data become hierarchical, and application can be for proximity and related searching, leading to nearest neighbour regression. But far more than that is ultrametric regression, taking into account the ultrametric topology associated with hierarchical structure. For our cognitive processes and analytics, ultrametric regression is how cognitive and analytical processing determines such system properties for regression purposes. By using contiguity-constrained, i.e. here, chronological, hierarchical clustering, then through multivariate time series, and changepoint analysis, hierarchy expresses changes at varying scales. At issue, quite generally, are multivariate time series. Furthermore, our partitioning of the chronologically constrained hierarchical clustering, so as to segment the multivariate time series, and determine changepoints, this is carried out using a wavelet transform in the ultrametric topological space. Th e case study here, of ultrametric wavelet regression of multivariate time series, is through application to Colombian confl ict analysis.