DSLs for Machine Learning
We are collaborating with Christopher Ré from Stanford University to improve the interface to the DeepDive machine learning system. More specifically, we aim to phrase the interface to DeepDive as an embedded domain-specific language with the eventual goal to integrate machine learning so closely and seamlessly into ordinary programs that it is as simple to define a function or data structure via machine learning than to define a function via ordinary programming. The goal of this thesis is to to make the first steps towards that goal. Concretely, the first step would consist in embedding DeepDive into a statically typed programming language such as Scala.
The ideal student for this project is a good (functional) programmer and is also knowledable in databases and machine learning. This is a research-oriented thesis project with the possibility of international collaboration and the perspective of publishing a successful thesis at a respected venue.