Software

StochasticSQP is a software package for solving optimization problems involving a stochastic objective and deterministic constraint functions. (Problems with deterministic objective functions can also be solved, although the code has not been optimized for that purpose.) StochasticSQP is written in Matlab.

• Link: https://github.com/frankecurtis/StochasticSQP
• Code written by: Albert S. Berahas, Frank E. Curtis, Daniel P. Robinson and Baoyu Zhou
• Related paper: Berahas, A. S., Curtis, F. E., Robinson, D., & Zhou, B. (2021). Sequential quadratic optimization for nonlinear equality constrained stochastic optimization. SIAM Journal on Optimization, 31(2), 1352-1379.

AggQN (Aggregated quasi-Newton) contains implementations of classes for storing and employing quasi-Newton (inverse) Hessian approximations for use in an optimization algorithm. An AggQN object that implements an aggregated BFGS strategy is currently available in Matlab. (An implementation in C++ is forthcoming.) The code is released under the MIT License.

• Link: https://github.com/frankecurtis/AggQN
• Code written by: Albert S. Berahas, Frank E. Curtis and Baoyu Zhou
• Related paper: Berahas, A. S., Curtis, F. E., & Zhou, B. (2022). Limited-memory BFGS with displacement aggregation. Mathematical Programming.

Graphical Lasso (GL) is a popular method for learning the structure of Gaussian graphical models, which is based on an L1 regularization technique. We develop a highly efficient approximate closed-form solution for Graphical Lasso. Our developed code is publicly available in Python and MATLAB.

• Link: https://github.com/AtomXT/GL-and-Thresholding
• Code written by: Tong Xu and Salar Fattahi
• Related paper: Fattahi, S., & Sojoudi, S. (2019). Graphical lasso and thresholding: Equivalence and closed-form solutions. Journal of machine learning research.

This is a software package for solving unconstrained optimization problems of the form, min f(x) = (1/n)*sum_i f_i(x). minSQN allows for the user to solve large-scale (sum-of-functions) optimization problems using one of 11 Stochastic Quasi-Newton methods.

• Link: https://github.com/keskarnitish/minSQN
• Code written by: Nitish Shirish Keskar and Albert S. Berahas