Proxy-Anchor and EVT-Driven Continual Learning Method for Generalized Category Discovery

Alireza Fathalizadeh, Roozbeh Razavi-Far · 2025

Read Proxy-Anchor and EVT-Driven Continual Learning Method for Generalized Category Discovery by Alireza Fathalizadeh, Roozbeh Razavi-Far (2025) 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-2504-08550.txt.

Description

Alireza Fathalizadeh and Roozbeh Razavi-Far tackle continual category discovery, where a vision system must keep spotting new classes in fresh data without forgetting old ones, in a 2025 paper. Sorting true novelties from known items is the crux, and extreme value theory has been the usual tool for drawing that line. Their method marries that theory to proxy anchors, reference points for each known class.

A probability rule around each anchor rejects unknown samples, and a new loss sharpens the embedding space. Extra machinery prunes surplus categories, while replay and distillation guard old knowledge. Experiments report gains over rival metric learning methods.

Catalog record

CREATORAlireza Fathalizadeh, Roozbeh Razavi-Far
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-08550

Browse the full library · About the Tower of Babel