A dependent type system for machine learning in Lisp
Current machine learning frameworks are built around relatively weak type systems. This is a problem because, at scale, machine learning applications are exceedingly intricate and computationally expensive, therefore making costly runtime errors unavoidable. This is where Vouivre comes into play. Using a dependent-type system, the project aims at enabling users to write machine-learning applications that solve real-world problems with compile-time validation of their correctness, thus preventing runtime errors at a reasonable computational cost.
This project was funded through theNGI0 Core Fund, a fund established byNLnet with financial support from the European Commission'sNext Generation Internet programme, under the aegis ofDG Communications Networks, Content and Technology under grant agreement No101092990.