Download the EDBT 2026 paper: A declarative, recursive SQL framework for composable machine learning ensembles
Modern hyperscale data warehouses store massive datasets that are increasingly used for machine learning, but building ensemble models often requires exporting data into external frameworks and managing complex, imperative pipelines. This paper presents Ocient’s SQL-native framework for composable machine learning ensembles, designed to let users define bagging, boosting, and stacking models declaratively within the Ocient Hyperscale Data Warehouse.
Presented at EDBT 2026 in Tampere, Finland, the paper details how Ocient enables heterogeneous and recursive ensemble pipelines that combine multiple OcientML model types, train directly inside the database, and reduce complex multi-stage workflows to a single SQL statement.