L-BFGS: Limited-memory BFGS and other memory-efficient Quasi-Newton optimization algorithms in R, with support for L1 (lasso) regularization. For machine learning applications and numerical optimization in high-dimensional settings.