to portfolio solver using a learning algorithm.

This paper presents a parallel Constraint Programming (CP) solver, based on the portfolio principle, which quickly solves constraint satisfaction and optimisation problems. The principle of the Portfolio method is to run N search strategies using N computing cores and the first strategy to satisfy the needs of the user stops all other strategies. In the usual portfolio principle, the number of search strategies is limited compared to the current number of computing cores used by the parallel machines. Using an internal parallelization for each search strategy, it is possible to run N search strategies using P computing cores with P > N ( N To P Portfolio ), as all the search strategies are scheduled using the same parallel framewok (Bobpp [1]). So, we can schedule dynamically the N search strategies there between in order to favour the strategy that finds a solution quickly and we give more cores to this privileged strategy. This Portfolio model is called the Adaptive Portfolio . The Adaptive Portfolio is used when the user wants to solve different CP problems and can not know the best strategy. In some industrial projects, such as the PAJERO project, we always solve different instances of the same CP problem. The novelty is to use the N To P Portfolio using a learning algorithm to predicts the number of cores to assign to each strategy automatically from a database that stores the ranks of all strategies used to solve a number of instances.