Show simple item record

dc.contributor.authorSheehan, Conor
dc.contributor.authorDay, Ben
dc.contributor.authorLiò, Pietro
dc.date.accessioned2019-05-27T23:30:17Z
dc.date.available2019-05-27T23:30:17Z
dc.date.issued2018-10-22
dc.identifier.urihttps://www.repository.cam.ac.uk/handle/1810/293144
dc.description.abstractOne-hot encoding is a labelling system that embeds classes as standard basis vectors in a label space. Despite seeing near-universal use in supervised categorical classification tasks, the scheme is problematic in its geometric implication that, as all classes are equally distant, all classes are equally different. This is inconsistent with most, if not all, real-world tasks due to the prevalence of ancestral and convergent relationships generating a varying degree of morphological similarity across classes. We address this issue by introducing curvature to the label-space using a metric tensor as a self-regulating method that better represents these relationships as a bolt-on, learning-algorithm agnostic solution. We propose both general constraints and specific statistical parameterizations of the metric and identify a direction for future research using autoencoder-based parameterizations.
dc.rightsAll rights reserved
dc.subjectcs.LG
dc.subjectcs.LG
dc.subjectstat.ML
dc.titleIntroducing Curvature to the Label Space
dc.typeArticle
prism.publicationNameCoRR
dc.identifier.doi10.17863/CAM.40293
rioxxterms.versionAM
rioxxterms.licenseref.urihttp://www.rioxx.net/licenses/all-rights-reserved
dc.contributor.orcidLio, Pietro [0000-0002-0540-5053]
rioxxterms.typeJournal Article/Review
cam.orpheus.success2021-06-09 No trace of acceptance or publication.
cam.orpheus.counter68
rioxxterms.freetoread.startdate2100-01-01


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record