Posted: 7th December 2018

The FDB project has been awarded an unrestricted research gift of USD 100,000 from Infor Corp (USA) to support on-going work on in-database machine learning. 

Professor Dan Olteanu of the Computer Science Department, the leader of the FDB project says: 'Our goal is to build a scalable system for training machine learning models over relational databases. We recently showed that popular machine learning models, including decision trees, factorization machines, and support vector machines can be trained directly inside a database engine by reformulating the training task as the problem of evaluating  batches of relational queries over the input database. Our approach comes with both theoretical and practical benefits. It enjoys lower computational complexity than the existing approaches, which means in practice training over larger datasets and orders-of-magnitude faster than state-of-the-art analytics systems.'

Dr. Kurt Stirewalt, Vice President of Software Development, Infor says: 'Infor is pleased to support the work of Professor Dan Olteanu on the FDB project, which addresses a fundamental obstacle to scaling machine learning to large data sets such as we encounter when forecasting demand in retail contexts.'