Few Shot Learning (FSL) or low shot learning is the learning process that trains the model to make predictions with the limited number of information. The main goal of few-shot learning is to build an accurate learning model with less training data. The importance of few-shot learning is test base learning, learning for rare cases, reducing data collection effort and computational costs. The few-shot learning models are categorized under three approaches based on similarity (e.g., matching network), learning (e.g., LSTMs), and data (e.g., generative models).

     Application fields of few-shot learning are computer vision, natural language processing, robotics, acoustic signal processing, Iot analytics, medical and mathematical applications. Some computer vision application tasks are Character recognition, Image and video classification, Object recognition, Object tracking, and Image segmentation. Few natural language processing application tasks are translation, sentiment classification, and multi-label text classification. Future advancements of few-shot learning are FSL with multi-modality data, meta learning-based FSL methods, transfer knowledge with FSL.