LSLOpt : An open‐source implementation of the step‐length controlled LSL‐BFGS algorithm

Abstract Numerical optimization is a common technique in various areas of computational chemistry, molecular modeling and drug design. It key element 3D techniques, for example, the protein–ligand poses small‐molecule conformers. Here, often BFGS algorithm or variants thereof are used. However, tends to make unreasonable large changes optimized system under certain circumstances. This behavior has been known long time different solutions have suggested. Recently, we analyzed our novel JAMDA scoring function detail proposed limited step length (LSL)‐BFGS as new solution problem excessively steps during optimization. The LSL‐BFGS allows control sizes Its unique feature inclusion arbitrary domain knowledge into selection sizes. introduce open‐source LSLOpt C++ library that implements this demonstrate its usage.