Class Incremental Learning
Class incremental learning (CIL) focuses on training machine learning models to continuously learn new classes of data without forgetting previously learned ones, a crucial challenge for real-world applications with evolving data streams. Current research emphasizes techniques like dynamic model architectures (e.g., adding task-specific adapters), generative replay methods to synthesize past data, and the use of pre-trained models to leverage existing knowledge. These advancements aim to improve accuracy and fairness while addressing issues like catastrophic forgetting and data imbalance, impacting fields such as medical image analysis, sound source localization, and personalized AI systems.
Papers
Less is more: Summarizing Patch Tokens for efficient Multi-Label Class-Incremental Learning
Thomas De Min, Massimiliano Mancini, Stéphane Lathuilière, Subhankar Roy, Elisa Ricci
Rethinking Class-Incremental Learning from a Dynamic Imbalanced Learning Perspective
Leyuan Wang, Liuyu Xiang, Yunlong Wang, Huijia Wu, Zhaofeng He
Delve into Base-Novel Confusion: Redundancy Exploration for Few-Shot Class-Incremental Learning
Haichen Zhou, Yixiong Zou, Ruixuan Li, Yuhua Li, Kui Xiao
Few-Shot Class Incremental Learning via Robust Transformer Approach
Naeem Paeedeh, Mahardhika Pratama, Sunu Wibirama, Wolfgang Mayer, Zehong Cao, Ryszard Kowalczyk
Brain-Inspired Continual Learning-Robust Feature Distillation and Re-Consolidation for Class Incremental Learning
Hikmat Khan, Nidhal Carla Bouaynaya, Ghulam Rasool
DynaMMo: Dynamic Model Merging for Efficient Class Incremental Learning for Medical Images
Mohammad Areeb Qazi, Ibrahim Almakky, Anees Ur Rehman Hashmi, Santosh Sanjeev, Mohammad Yaqub
Toward industrial use of continual learning : new metrics proposal for class incremental learning
Konaté Mohamed Abbas, Anne-Françoise Yao, Thierry Chateau, Pierre Bouges
Multi-Label Continual Learning for the Medical Domain: A Novel Benchmark
Marina Ceccon, Davide Dalle Pezze, Alessandro Fabris, Gian Antonio Susto