Of Dice and Games: A Theory of Generalized Boosting

Marco Bressan, Nataly Brukhim, Nicolò Cesa-Bianchi, Emmanuel Esposito, Yishay Mansour, Shay Moran, Maximilian Thiessen · 2024

Read Of Dice and Games: A Theory of Generalized Boosting by Marco Bressan, Nataly Brukhim, Nicolò Cesa-Bianchi, Emmanuel Esposito, Yishay Mansour, Shay Moran, Maximilian Thiessen (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-08012.txt.

Description

Marco Bressan and six co-authors build a theory of generalized boosting, in work from 2024. Boosting theory has mostly dealt with the plain 0-1 loss, where all errors count the same. Real tasks often price errors in distinct ways, as in medical diagnosis, and may track several goals at once, which calls for cost-sensitive and multi-objective losses.

Their theory sorts weak learning guarantees into trivial, boostable and middle classes. For binary tasks, a clean split holds: each guarantee is either trivial or boostable. Multi-class tasks show a richer middle ground. A geometric view of boosting drives the results and reveals a surprise link between the two loss types.

Catalog record

CREATORMarco Bressan, Nataly Brukhim, Nicolò Cesa-Bianchi, Emmanuel Esposito, Yishay Mansour, Shay Moran, Maximilian Thiessen
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE142 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-08012

Browse the full library · About the Tower of Babel