NGAME: Negative Mining-aware Mini-batching for Extreme Classification
This work improves efficiency and accuracy for extreme classification tasks, such as tagging data with millions of labels, which is incremental but impactful for applications like personalized ads in search engines.
The paper tackles the challenge of training deep extreme classification models with large encoder architectures by addressing memory overheads that limit mini-batch sizes, introducing NGAME, a mini-batch creation technique that enables larger batches and achieves up to 16% higher accuracy on benchmarks and 23% gains in click-through-rates in live tests.
Extreme Classification (XC) seeks to tag data points with the most relevant subset of labels from an extremely large label set. Performing deep XC with dense, learnt representations for data points and labels has attracted much attention due to its superiority over earlier XC methods that used sparse, hand-crafted features. Negative mining techniques have emerged as a critical component of all deep XC methods that allow them to scale to millions of labels. However, despite recent advances, training deep XC models with large encoder architectures such as transformers remains challenging. This paper identifies that memory overheads of popular negative mining techniques often force mini-batch sizes to remain small and slow training down. In response, this paper introduces NGAME, a light-weight mini-batch creation technique that offers provably accurate in-batch negative samples. This allows training with larger mini-batches offering significantly faster convergence and higher accuracies than existing negative sampling techniques. NGAME was found to be up to 16% more accurate than state-of-the-art methods on a wide array of benchmark datasets for extreme classification, as well as 3% more accurate at retrieving search engine queries in response to a user webpage visit to show personalized ads. In live A/B tests on a popular search engine, NGAME yielded up to 23% gains in click-through-rates.