arXiv:2504.20566v2 Announce Type: replace 
Abstract: Online class-incremental learning (OCIL) focuses on gradually learning new classes (called plasticity) from a stream of data in a single-pass, while concurrently preserving knowledge of previously learned classes (called stability). The primary challenge in OCIL lies in maintaining a good balance between the knowledge of old and new classes within the continually updated model. Most existing methods rely on explicit knowledge interaction through experience replay, and often employ exclusive training separation to address bias problems. Nevertheless, it still remains a big challenge to achieve a well-balanced learner, as these methods often exhibit either reduced plasticity or limited stability due to difficulties in continually integrating knowledge in the OCIL setting. In this paper, we propose a novel replay-based method, called Balanced Inclusive Separation for Online iNcremental learning (BISON), which can achieve both high plasticity and stability, thus ensuring more balanced performance in OCIL. Our BISON method proposes an inclusive training separation strategy using dual classifiers so that knowledge from both old and new classes can effectively be integrated into the model, while introducing implicit approaches for transferring knowledge across the two classifiers. Extensive experimental evaluations over three widely-used OCIL benchmark datasets demonstrate the superiority of BISON, showing more balanced yet better performance compared to state-of-the-art replay-based OCIL methods.

قدمت دراسة جديدة طريقة تُسمى Balanced Inclusive Separation for Online Incremental learning (BISON)، والتي تتناول تحديات التعلم التدريجي عبر الإنترنت (OCIL) من خلال تحقيق التوازن بين الاحتفاظ بالمعرفة من الفئات القديمة أثناء دمج الفئات الجديدة. تهدف هذه الطريقة إلى تحسين كل من المرونة والثبات في نماذج التعلم الآلي، مما يمثل تقدماً كبيراً في هذا المجال.

Un nuevo estudio ha presentado un método llamado Balanced Inclusive Separation for Online Incremental learning (BISON), que aborda los desafíos del aprendizaje incremental en línea (OCIL) al equilibrar la retención de conocimientos de clases antiguas mientras se integran nuevas. Este método busca mejorar tanto la plasticidad como la estabilidad en los modelos de aprendizaje automático, lo que representa un avance significativo en el campo.

Une nouvelle étude a introduit une méthode appelée Balanced Inclusive Separation for Online Incremental learning (BISON), qui aborde les défis de l'apprentissage incrémental en ligne (OCIL) en équilibrant la rétention des connaissances des anciennes classes tout en intégrant de nouvelles. Cette méthode vise à améliorer à la fois la plasticité et la stabilité des modèles d'apprentissage automatique, une avancée significative dans le domaine.

A new study has introduced a method called Balanced Inclusive Separation for Online Incremental learning (BISON), which addresses the challenges of online class-incremental learning (OCIL) by balancing the retention of knowledge from old classes while integrating new ones. This method aims to enhance both plasticity and stability in machine learning models, a significant advancement in the field.

Balanced Online Class-Incremental Learning via Dual Classifiers

One More Thing in AI – Your Shortcut to AI Mastery

Balanced Online Class-Incremental Learning via Dual Classifiers

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

ClassX

ClassroomIO

Octofy

CodeSpaced

Ready to build your own newsroom?