Anomaly Sample

Anomaly sample detection focuses on identifying data points deviating significantly from a norm, typically using only normal data for training. Current research emphasizes unsupervised and semi-supervised approaches, employing diverse model architectures including autoencoders, diffusion models, transformers, and graph neural networks, often enhanced by techniques like contrastive learning, pseudo-anomaly generation, and knowledge integration. This field is crucial for various applications, from industrial quality control and network security to medical diagnosis, with advancements driving improved accuracy and efficiency in detecting anomalies across diverse data types.