We are proud to release the SUMO Toolbox 2016b. This new release can be downloaded through the download page.
- Added support for classification problems for FLOLA-Voronoi
- Added expression generator for SVM and LS-SVM (does not cover all cases, regression only)
- Added knowledge-gradient optimization for stochastic simulation optimization.
- Added EnsembleRanker to calculate the mean of a ranker criterion (e.g., for slice sampling the EI)
- Used in combination with the numSlices option in KrigingFactory and GaussianProcessFactory
- Added numSlices option to the EGO ModelBuilder
- Used with the EGO ModelBuilder to drive model selection of expensive models
- Added support for mixed integer optimization problems for MatlabGA optimizer
- Kriging models can predict the full posterior covariance matrix now (predictCovarianceMatrix)
- Added Predictive Entropy Search for Bayesian optimization (experimental)