Today there is only little support for developing softwarethat incorporates a machine learning or a data mining component. Toalleviate this situation, we propose to develop programming languagesfor machine learning and data mining. We also argue that such languagesshould be declarative and should be based on constraint programmingmodeling principles. In this way, one could declaratively specify the problemof machine learning or data mining problem of interest in a high-levelmodeling language and then translate it into a constraint satisfactionor optimization problem, which could then be solved using particularsolvers. These ideas are illustrated on problems of constraint-based itemsetand pattern set mining.
展开▼
