Class-Incremental Learning with Repetition
Type
conference paper
Date Issued
2023
Author(s)
;
Andrea Cossu
;
Antonio Carta
;
Julio Hurtado
;
Lorenzo Pellegrini
;
Davide Bacciu
;
Vincenzo Lomonaco
;
Editor(s)
Sarath Chandar
Razvan Pascanu
Hanie Sedghi
Doina Precup
Abstract
Real-world data streams naturally include the repetition of previous concepts. From a Continual Learning (CL) perspective, repetition is a property of the environment and, unlike replay, cannot be controlled by the agent. Nowadays, the Class-Incremental (CI) scenario represents the leading test-bed for assessing and comparing CL strategies. This scenario type is very easy to use, but it never allows revisiting previously seen classes, thus completely neglecting the role of repetition. We focus on the family of Class-Incremental with Repetition (CIR) scenario, where repetition is embedded in the definition of the stream. We propose two stochastic stream generators that produce a wide range of CIR streams starting from a single dataset and a few interpretable control parameters. We conduct the first comprehensive evaluation of repetition in CL by studying the behavior of existing CL strategies under different CIR streams. We then present a novel replay strategy that exploits repetition and counteracts the natural imbalance present in the stream. On both CIFAR100 and TinyImageNet, our strategy outperforms other replay approaches, which are not designed for environments with repetition.
Language
English
Event Title
Conference on Lifelong Learning Agents
File(s)![Thumbnail Image]()
open.access
Name
hemati23b.pdf
Size
1.44 MB
Format
Adobe PDF
Checksum (MD5)
a57fddeb8bb836ea8ad4fd2da34f5188