A Survey of Few-Shot Learning Research Based on Deep Neural Network

Abstract

With the successful development of deep learning techniques in recent years, deep neural networks have achieved excellent results in both computer vision and natural language processing by relying on large-scale datasets but still face significant challenges in solving the problem of learning from few-shot. Inspired by the ability of humans to learn to recognize objects as a way to simulate the cognitive process of learning from a small sample size, few-shot learning is a hot topic of research in deep neural networks today. It is also a significant and challenging problem. This paper first introduces the research background and definition of few-shot learning, introduces the relevant models, and summarizes and analyzes the common approaches to the problem of few-shot learning based on deep neural networks at the present stage, which are divided into four types: data augmentation, model fine-tuning, metric learning and meta-learning. Finally, popular datasets for few-shot learning are described, the paper is concluded and future research directions are discussed.

Fine-tuning pre-trained neural networks for medical image classification in small clinical datasets

“…This study supports research on health informatics by addressing and assessing an intelligent method that can learn from small labeled datasets from distinct domains. These datasets illustrate many real-world scenarios in which it is hard to find a reasonable amount of labeled images for infrequent diseases or conditions [73], often leading to imbalanced class distribution. The best deep learning settings investigated in this work performed competitively, offering an alternative to classify medical image examinations with a relatively low number of abnormal and normal samples.…”

“…This study can be linked to the few-shot learning paradigm, a maturing subarea of machine learning that focuses on enabling models to rapidly learn and generalize from a limited number of labeled examples [73]. Although researchers have investigated distinct techniques and frameworks to apply few-shot learning in small clinical datasets [39,55,65], our proposal differentiates from them by employing transfer learning via several FT settings in two domains.…”