Saved in:
Bibliographic Details
Main Authors: Amin, Md Hasibul, Alam, Tamzid Tanvi
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.07648
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911205565661184
author Amin, Md Hasibul
Alam, Tamzid Tanvi
author_facet Amin, Md Hasibul
Alam, Tamzid Tanvi
contents Continual learning the ability of a neural network to learn multiple sequential tasks without catastrophic forgetting remains a central challenge in developing adaptive artificial intelligence systems. While deep learning models achieve state-of-the-art performance across domains, they remain limited by overfitting and forgetting. This paper introduces Cluster-Aware Replay (CAR), a hybrid continual learning framework that integrates a small, class-balanced replay buffer with a regularization term based on Inter-Cluster Fitness (ICF) in the feature space. The ICF loss penalizes overlapping feature representations between new and previously learned tasks, encouraging geometric separation in the latent space and reducing interference. Using the standard five-task Split CIFAR-10 benchmark with a ResNet-18 backbone, initial experiments demonstrate that CAR better preserves earlier task performance compared to fine-tuning alone. These findings are preliminary but highlight feature-space regularization as a promising direction for mitigating catastrophic forgetting.
format Preprint
id arxiv_https___arxiv_org_abs_2510_07648
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Continual Learning for Adaptive AI Systems
Amin, Md Hasibul
Alam, Tamzid Tanvi
Machine Learning
Continual learning the ability of a neural network to learn multiple sequential tasks without catastrophic forgetting remains a central challenge in developing adaptive artificial intelligence systems. While deep learning models achieve state-of-the-art performance across domains, they remain limited by overfitting and forgetting. This paper introduces Cluster-Aware Replay (CAR), a hybrid continual learning framework that integrates a small, class-balanced replay buffer with a regularization term based on Inter-Cluster Fitness (ICF) in the feature space. The ICF loss penalizes overlapping feature representations between new and previously learned tasks, encouraging geometric separation in the latent space and reducing interference. Using the standard five-task Split CIFAR-10 benchmark with a ResNet-18 backbone, initial experiments demonstrate that CAR better preserves earlier task performance compared to fine-tuning alone. These findings are preliminary but highlight feature-space regularization as a promising direction for mitigating catastrophic forgetting.
title Continual Learning for Adaptive AI Systems
topic Machine Learning
url https://arxiv.org/abs/2510.07648