Self-Adversarial Negative Sampling

Self-Adversarial Negative Sampling is a technique used in natural language processing to improve the efficiency of negative sampling in methods like word embeddings and knowledge graph embeddings. Negative sampling is a process that involves the sampling of negative triplets that are false in order to provide meaningful information during training. However, traditional negative sampling samples negatives uniformly, which leads to inefficiencies since many samples are blatantly false. This is whe