LOCALSOLVER math solver 10.0 version released: welcome to apply for trial experience
LocalSolver has achieved gratifying results in 2020. More commercial customers have joined the local solver user family, mainly Bosch Automotive (Germany, American company), CEZ Distribuce Power Grid ( Czech Slovakia ) , NEWTON Vaureal (France), Sony (Japan) .
LocalSolver 10.0 version in 2021 Nian 1 Yue 18 smooth release date, adds many new features and solving solving performance improvements:
- VRP path optimization — There are good solutions to the problem of multiple VRP variants:
Faster preprocessing of large instances ( >10000 customers); better solutions to the time window problem: 60% of large instances (over 400 customers) have been improved, the average of Solomon ( CVRPTW ) and Li&Lim instances ( PDPTW ) The improvement rate is 5% ; a better solution to the non-uniform fleet problem; calculation of the lower bound of the vehicle routing problem: In 1000 cities, the average gap of TSP instances is 2.3% , and the average gap of CVRP instances is 2.6% , up to 100 Customers; calculation of the lower bound of the quality of vehicle routing problems (based on Held-Karp Lagrangian relaxation).
- Packaging and clustering problems : a better solution to the problem of heterogeneous containers ; clustering problems better solve geometric clustering problems, such as K- means, extended to 10,000 points. The optimality gap is less than 5% within 1 minute .
- Black box optimization:
Better decision-making model of integer solution - to improve our internal benchmarks than 20% ; the Oracle number of calls decreased by 33%
Analyze constraints: a better solution than the most advanced black box solver that handles constraints; fewer orcale calls.
- Global optimization:
Lexicon multi-objective problems: better solutions and boundaries for nonlinear models; better cuts for linear models; positive effects on our 15% internal benchmark.
Performance improvement for the sum of squares problem: faster convergence on the scale problem.
- Localsolver programming language ( LSP )
The modeler API of Python , java , C++ , and C++ introduces the LSP model in production ; the standard LSP library provides JSON parsing
( 6 ) The latest version introduces the surrogate modeling function: add analysis constraints to the black box model, as in the classic model.
Contact the LOCALSOLVER Chinese agent Wuxi Xunhe Information Technology Co., Ltd. to obtain the trial version of version 10.0.
