Why Two Sigma is using SigOpt for Automated Parameter Tuning

This post details the integration of SigOpt's optimization service into Two Sigma's research platform. It highlights the challenges of parameter tuning for 'black-box' functions in simulations and ML algorithms, contrasting grid search and Bayesian optimization with SigOpt's performance. Key benefits discussed include significantly faster tuning times (e.g., 24 days to 3 days for an ML algorithm), the ability to perform multi-metric optimization to explore Pareto frontiers, and efficient utilization of compute resources through asynchronous parallelization.