June 4, 2019
pSeven 6.15.1 Release
DATADVANCE development team announces the release of the pSeven 6.15.1 update, which further enhances user experience when building and evaluating approximation models. It contains several internal improvements to the MoA technique and adds the initial model support for more approximation techniques in the Model builder tool.
In this release, we mainly focused on improving the Mixture of Approximators (MoA) technique. Due to the changes in internal algorithms, it became faster and more accurate in many cases. Other improvements touch on the extended initial model support for the Model builder tool in Analyze.
As you know, an advanced training mode which uses an already trained (initial) model as a “starting point” and trains a new model which updates or improves the initial one was first introduced in pSeven 6.8 and was initially supported only by the Gradient-Boosted Regression Trees (GBRT) technique. pSeven 6.15.1 adds the initial model support also for the Mixture of Approximators (MoA) and High-Dimensional Approximation with Gaussian Processes (HDAGP) techniques in the manual mode. Actually, it means that the mentioned techniques can use the initial model for various purposes, namely: to update it with new data, improve model accuracy or save training time.
pSeven 6.15.1 is a recommended update for pSeven users who build and manage approximation models using the Predictive Modeling tools to solve various engineering tasks. This update also contains all recent enhancements introduced since pSeven 6.15. If you are planning an upgrade from a non-recent version of pSeven, it is recommended to update directly to pSeven 6.15.1, skipping the previous versions.