February 10, 2020
pSeven 6.16.2 Release
DATADVANCE development team announces the release of the pSeven 6.16.2 update. This release adds experimental support for exporting approximation models to C# library format, applies a few minor updates to predictive modeling tools, and fixes an important issue with models trained to keep linear dependencies between outputs.
Following the support for exporting approximation models to C# source code added in 6.16.1, pSeven can now automatically compile a model from C# source to export a .NET DLL. This feature uses a system C# compiler, so it requires a system package which provides the compiler - for example, .NET Framework for Windows or .NET Core SDK for Linux. The new model export format is supported by the ApproxBuilder and Approximation model blocks, as well as by predictive modeling tools in Analyze. Model export steps are generally the same as for other formats: see block documentation and section Model Import and Export in the pSeven user manual. Note that exporting models to C# source and DLL formats requires an up to date license file valid for pSeven 6.16 and above, since older licenses do not enable this feature.
This update also resolves an issue where the trained model could show unexpectedly high errors for some outputs, if the search for linear dependencies between outputs is enabled in training settings. This issue appeared due to certain properties of training data, so many models were not affected. However you should consider updating the models trained in previous versions of pSeven, especially if you are using the linear output dependency feature.
Additionally, pSeven 6.16.2 contains a few minor improvements in predictive modeling tools and several bugfixes, which you can find described in the release changelog. It is a recommended update, which includes all enhancements introduced in pSeven 6.16 and later. If you are planning to update from an earlier version of pSeven, it is also recommended to skip previous versions and update directly to pSeven 6.16.2.