Also known as memory-based learning
family of learning algorithms that, instead of performing explicit generalization, compares new problem instances with instances seen in training, which have been stored in memory
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).