632 KB0 KB200220242023Kastrin, Andrejštevilka:1/2letnik:28str. 16-23ISSN:1318-2145COBISSID_HOST:179058947URN:URN:NBN:SI:doc-BND2159SslSlovensko društvo za medicinsko informatikoInformatica Medica Slovenicahigh-dimensional datamachine learningmnogorazsežni podatkistatisticsstatistikastrojno učenjeO nekaterih lastnostih mnogorazsežnih podatkovij| On some properties of high-dimensional data sets|The extensive use of high-dimensional data to examine certain research phenomena has expanded in parallel with high-throughput technologies in various scientific fields. However, several statistical challenges arise when analysing high-dimensional datasets, such as low identifiability, numerical instability, and overfitting. Before delving into the complex analysis of high-dimensional data, a solid foundation of their inherent properties is crucial. This review aims to illustrate the geometric properties inherent in the statistical analysis of high-dimensional data by examining the behaviour of hypercubes and hyperspheres in a high-dimensional context. An overview of literature is also provided to guide the students in further studyTEXTznanstveno časopisjejournalsSlovenian National E-content AggregatorNational and University Library of SloveniaSlovensko društvo za medicinsko informatiko