{"pk":25705,"title":"Generating Hyperdimensional Distributed Representations from Continuous-\nValued Multivariate Sensory Input","subtitle":null,"abstract":"Hyperdimensional computing (HDC) refers to the\nrepresentation and manipulation of data in a very high\ndimensional space using random vectors. Due to the high\ndimensionality, vectors of the space can code large amounts\nof information in a distributed manner, are robust to variation,\nand are easily distinguished from random noise. More\nimportantly, HDC can be used to represent compositional and\nhierarchical relationships and recursive operations between\nentities using fixed-size representations, making it intriguing\nfrom a cognitive modeling point of view. However, the\nmajority of the existing work in this area has focused on\nmodeling discrete categorical data. This paper presents a new\nmethod for mapping continuous-valued multivariate data into\nhypervectors, enabling construction of compositional\nrepresentations from non-categorical data. The mapping is\nstudied in a word classification task, showing how rich\ndistributed representations of spoken words can be encoded\nusing HDC-based representations","language":"eng","license":{"name":"","short_name":"","text":null,"url":""},"keywords":[{"word":"hyperdimensional computing; distributed\nrepresentations; speech recognition; memory"}],"section":"Papers","is_remote":true,"remote_url":"https://escholarship.org/uc/item/3db1b0x6","frozenauthors":[{"first_name":"Okko","middle_name":"","last_name":"Rasanen","name_suffix":"","institution":"Aalto University","department":""}],"date_submitted":null,"date_accepted":null,"date_published":"2015-01-01T18:00:00Z","render_galley":null,"galleys":[{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/25705/galley/15329/download/"}]}