We are very happy to announce that CPLEX Optimization Studio 12.10 was released last Friday.
This version includes:
- Ability for CP Optimizer users to programmatically monitor the solve and stop it ahead of time if appropriate.
- Ability for CPLEX users to specify their own branching strategy in the new callback framework, for additional performance using problem-specific knowledge.
- Much increased stability of the OPL runtime, which four times as many items resolved this year than over the six previous years combined.
We also introduced new algorithms to improve the speed of the solvers. The most notable improvements are:
- CPLEX is the first commercial solver to implement a Machine Learning based algorithm to make automatic decisions over some algorithmic choices. This gives a 1.8x speedup on solving Mixed-Integer Quadratic Problems.
- CPLEX is also the only commercial solver to offer an automatic Benders decomposition algorithm. Thanks to several algorithmic improvements, version 12.10 is 1.8x faster than previous version.
- For CP Optimizer, version 12.10 builds on the excellent 12.9 to get a combined 5x speedup on feasibility combinatorial problems, where the goal is to prove the existence of a solution as quickly as possible, as well as 1.5x speedup on Scheduling problems.
- On Mixed-Integer Problems, performance in 12.10 improved by 10% on average.