New publication in the journal Pattern Recognition Letters

A team of researchers from Universidad Austral de Chile (Poblete, Aillapi, Duarte, Veas and Otondo) studied how to best classify data — sounds, images — once artificial intelligence has compressed them into a compact mathematical representation. They compared four classification methods across eight different types of data (from Chilean amphibian calls to everyday images) and found that no single method performs best in every case: the ideal choice depends on how the data within each category are internally organised.

Pattern Recognition Letters is a peer-reviewed scientific journal published by Elsevier (since 1982), specialising in artificial intelligence, pattern recognition and machine learning.

You can access the publication here: