A MIP Backend for the IDP System
This work addresses a specific optimization issue for users of knowledge base systems, but it is incremental as it builds on existing backend comparisons.
The paper tackled the problem of determining when a Mixed Integer Programming (MIP) solver is preferable to a Constraint Programming (CP) solver as a backend for the IDP knowledge base system, and it reported experimental comparisons between CPLEX and MiniSAT(ID) backends.
The IDP knowledge base system currently uses MiniSAT(ID) as its backend Constraint Programming (CP) solver. A few similar systems have used a Mixed Integer Programming (MIP) solver as backend. However, so far little is known about when the MIP solver is preferable. This paper explores this question. It describes the use of CPLEX as a backend for IDP and reports on experiments comparing both backends.