SigOpt is a proud sponsor of and contributor to QMCPy, an open source project and community dedicated to making quasi-Monte Carlo (QMC) methods more accessible that was launched last month.
SigOpt’s mission is to accelerate and amplify the impact of modelers everywhere. In pursuit of this mission, we look for opportunities to give back to the broader community of researchers, scientists, and engineers so they can be empowered in their own work. This includes the SigOpt Academic Program, which offers free access to SigOpt for nonprofits, labs, and universities that aim to publish their research. It includes a SigOpt for Good program that extends free access to SigOpt to anyone – enterprises or academic institutions alike – who are using SigOpt to accomplish a public goal, such as research related to COVID-19.
It also includes collaboration on and financial support for open source projects related to computational mathematics/statistics, machine learning, or platform engineering. It is with this purpose in mind that we contributed to and supported the QMCPy project.
Why did we choose to contribute to this project? SigOpt’s software includes a proprietary ensemble of Bayesian and global optimization algorithms that our customers and users apply to tune parameters or hyperparameters for any model. QMC is a valuable component in the computation of key quantities during this optimization process, allowing us to run more accurate computations in the same amount of time.
In conjunction with the announcement of QMCPy, the project leaders ran a series of blog posts explaining QMC, why they developed QMCPy, and how to use QMCPy, including a basic explanation of low discrepancy sequences. We will cross-post these articles to promote their work.
Thank you to our main collaborators in this project, Fred Hickernell and Aleksei Sorokin at the Illinois Institute of Technology and Sou-Cheng Choi at Kamakura Corporation, who were the driving force behind QMCPy. Learn more about our research, join our Academic Program, try our enterprise product, and, most importantly, try out QMCPy.