Published 2024-12-30
How to Cite
This work is licensed under a Creative Commons Attribution 4.0 International License.
Abstract
The pursuit of Artificial General Intelligence (AGI) necessitates models capable of rapid adaptation to novel tasks with minimal data, akin to human learning. This paper introduces TSMB (TWO-STAGE MANIFOLD-BASED FEW-SHOT LEARNING), a novel approach to Few-Shot Learning (FSL) that leverages unsupervised learning to harness the geometric distribution of data across tasks. TSMB refines feature representations through a two-stage process: first, by leveraging the topological structure of high-dimensional data to fine-tune general features, and second, by generating virtual samples to integrate semi-supervised learning. This method aims to address the challenges of data scarcity and overfitting, common in FSL. The model comprises a backbone network for feature extraction, a manifold learning module to capture topological results, a manifold support point section to assist learning, and a denoising prototype classifier for decision-making. TSMB demonstrates enhanced generality and performance, offering a promising direction for advancing FSL towards more ethical, sustainable, and effective AI applications.