MatrixNet is a proprietary machine learning algorithm developed by Yandex and used widely throughout the company products. The algorithm is based on gradient boosting, and was introduced since 2009.