Semantics-based selection and code bloat reduction techniques for genetic programming

1. Three semantic tournament selection are proposed, including TS-R, TS-S and TS-P. For further improvements, the best method, TS-S is combined with RDO, and the resulting method is called TS-RDO.
2. A novel semantic approximation technique (SAT) is proposed. Besides, two other versions of SAT are also introduced.
3. Two methods, SA and DA based on semantic approximation technique for reducing GP code bloat are proposed. Additionally, three other bloat control methods based on the variants of SAT, including SAT-GP, SAS- GP and PP-AT are introduced.