1: \begin{abstract}

2: 

3: The search strategy of a CP solver is determined by the variable and value ordering heuristics it employs and by the branching scheme it follows.

4: Although the effects of variable and value ordering heuristics on search effort have been widely studied, the effects of different branching schemes have received less attention. %The standard branching scheme employed by the majority of modern finite domain CP solvers is 2-way branching. An alternative branching scheme that has  been extensively used in the past and is still used by some solvers is $d$-way branching. Domain splitting, originating from numerical CSPs, is also applied in some cases, especially on optimization problems. Set branching methods where branching is performed on groups of values have also been proposed but have not gained wide appeal.

5: In this paper we study this effect %of different branching schemes on search effort

6: through an experimental evaluation that includes

7: standard branching schemes such as 2-way, $d$-way, and dichotomic domain splitting, as well as variations of set branching where branching is performed on sets of values. We also propose and evaluate a generic approach to set branching where the partition of a domain into sets is created using the scores assigned to values by a value ordering heuristic, and a clustering algorithm from machine learning. Experimental results %on a wide variety of benchmarks

8: demonstrate that although exponential differences between branching schemes, as predicted in theory between 2-way $d$-way branching, are not very common, still the choice of branching scheme can make quite a difference on certain classes of problems. Set branching methods are very competitive with 2-way branching and outperform it on some problem classes. A statistical analysis of the results reveals that our generic clustering-based set branching method is the best among the methods compared.

9: 

10: %, sometimes by a large margin.

11: 

12: \end{abstract}