ML4all is a machine learning (ML) system that aims at abstracting ML algorithms. It provides an easy-to-use interface purely based on user defined functions (UDFs). Users can express most of their ML tasks via this abstraction. ML4all can also take care of choosing the right Gradient Descent (GD) algorithm to run users tasks, e.g., logistic regression or SVMs. It leverages Rheem to plug physical optimizations and run ML tasks on top of any data processing platform. Thanks to this, ML4all can offer an easy-to-use ML task specification and provide orders of magnitude better performance than baseline systems. The source code of ML4all’s abstraction is available here

Back to projects