One Class Classification

One-class classification (OCC) focuses on building models that can identify data points belonging to a single, known class, without requiring examples of the other classes. Current research emphasizes developing efficient algorithms, such as those based on linear-time operations or subspace learning, and addressing challenges like robustness to adversarial examples and the effective use of limited data, often employing models like Support Vector Data Description (SVDD) and deep learning architectures (e.g., autoencoders, transformers). OCC finds significant application in anomaly detection across diverse fields, including medical diagnosis, cybersecurity, and industrial quality control, offering solutions where obtaining comprehensive labeled datasets is difficult or impossible.