arXiv:2511.21490v1 Announce Type: cross 
Abstract: We present a novel training approach, named Merge-and-Bound (M&B) for Class Incremental Learning (CIL), which directly manipulates model weights in the parameter space for optimization. Our algorithm involves two types of weight merging: inter-task weight merging and intra-task weight merging. Inter-task weight merging unifies previous models by averaging the weights of models from all previous stages. On the other hand, intra-task weight merging facilitates the learning of current task by combining the model parameters within current stage. For reliable weight merging, we also propose a bounded update technique that aims to optimize the target model with minimal cumulative updates and preserve knowledge from previous tasks; this strategy reveals that it is possible to effectively obtain new models near old ones, reducing catastrophic forgetting. M&B is seamlessly integrated into existing CIL methods without modifying architecture components or revising learning objectives. We extensively evaluate our algorithm on standard CIL benchmarks and demonstrate superior performance compared to state-of-the-art methods.

تم تقديم نهج تدريبي جديد يسمى Merge-and-Bound (M&B) للتعلم التدريجي للفئات (CIL)، والذي يركز على التلاعب المباشر بأوزان النموذج في فضاء المعلمات. تتضمن الطريقة دمج الأوزان بين المهام وداخل المهام لتعزيز التعلم مع تقليل النسيان الكارثي، مما يسمح بتحسين فعال للنماذج دون تعديل الهياكل الحالية.

Se ha introducido un nuevo enfoque de entrenamiento llamado Merge-and-Bound (M&B) para el Aprendizaje Incremental por Clases (CIL), que se centra en la manipulación directa de los pesos del modelo en el espacio de parámetros. El método incorpora la fusión de pesos inter-tarea e intra-tarea para mejorar el aprendizaje mientras minimiza el olvido catastrófico, permitiendo una optimización efectiva de los modelos sin alterar las arquitecturas existentes.

Une nouvelle approche d'entraînement nommée Merge-and-Bound (M&B) a été introduite pour l'apprentissage incrémental par classe (CIL), se concentrant sur la manipulation directe des poids du modèle dans l'espace des paramètres. La méthode intègre la fusion de poids inter-tâches et intra-tâches pour améliorer l'apprentissage tout en minimisant l'oubli catastrophique, permettant une optimisation efficace des modèles sans modifier les architectures existantes.

A novel training approach named Merge-and-Bound (M&B) has been introduced for Class Incremental Learning (CIL), focusing on direct manipulation of model weights in the parameter space. The method incorporates inter-task and intra-task weight merging to enhance learning while minimizing catastrophic forgetting, allowing for effective model optimization without altering existing architectures.

Merge and Bound: Direct Manipulations on Weights for Class Incremental Learning

One More Thing in AI – Your Shortcut to AI Mastery

Merge and Bound: Direct Manipulations on Weights for Class Incremental Learning

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

Https

LCW

FastML

Hypertune

Ready to build your own newsroom?