A Comprehensive Review on AutoMeta: An Adapti...

A Comprehensive Review on AutoMeta: An Adaptive Meta-Learning System for Dynamic Task Shifts

Author Name : Amith Subodh, Aniket M H, B G Parinith, B H Adithya Parashara, M. S. Prapulla Kumar

DOI: https://doi.org/10.56025/IJARESM.2025.1305253070

ABSTRACT In fields where labeled data is limited, meta-learning has become increasingly popular as a solution to few-shot learn- ing challenges. The dynamic nature of real-world clinical data in medical imaging causes frequent changes in task allocation, which traditional meta-learning techniques frequently ignore. This project presents AutoMeta, an adaptive meta-learning system that can dynamically adjust itself based on past task knowledge and detect changes in task distribution. In contrast to conventional MAML, AutoMeta improves performance on chang- ing few-shot medical tasks by combining contrastive adaptation, task embeddings, change-point detection, and a memory bank. Standard medical imaging datasets like HAM10000 and ISIC2018 are used to assess the system’s performance on lesion classifi- cation tasks. The suggested approach exhibits robust learning in dynamically changing contexts, enhanced generalization, and quicker adaption.