November 20, 2020
pSeven 6.17 Release
DATADVANCE development team announces the availability of pSeven 6.17, a new version of our flagship platform for data analysis and optimization. This is a major release bringing notable enhancements to design exploration, integration and predictive modeling tools. Read on for more information on the release highlights.
Automation & Integration
One of the major features of this release is the possibility to create User Blocks. pSeven provides a full-featured block template, which you can use to extend the pSeven’s block library with new integrations.
A new custom block generated from the template works out-of-the-box: it already contains a graphical interface, where the user can select parameters to create input and output ports — similarly to other integration type blocks, and also supports error handling, sandbox directories, and other common block features. The amount of required custom code is minimal: the block just needs 2 Python functions, which build the tree of parameters and process the input data, to wrap any software tool you need, add a graphical interface to an existing Python script, add support for a new model format, and so on.
Another notable improvement is related to STAR-CCM+ Block. The block now supports all STAR-CCM+ licensing schemas, including the support for Lite and Power licenses.
Design Exploration
Significant updates to the Design Space Exploration block are:
- Support of stepped variables. A stepped variable sticks to the grid defined by the variable’s bounds and step size. Stepped variables are intended for the common case of a continuous quantity (geometry parameter, load, power), which can be changed only in certain increments due to various practical reasons.
- Hint “Resolution” added to DSE block to specify resolution for variables. Specifying resolution for variables essentially sets precision of optimization techniques, adjusting their algorithms for a particular task. Compared to running optimization with maximum numerically possible resolution, setting proper resolution typically results in improved runtime (less response evaluations) at the cost of insignificant change in solution quality.
- Significant increase of the Orthogonal array technique performance and stability for big samples and designs where variables have many levels (about 10 or more).
- Improved Latin hypercube sampling technique results quality when it is used with an initial sample.
- Uncertainty quantification block improvement: reports generated by the block now contain more details on analysis variables, including the type and parameters of their distributions.
Predictive Modeling
Predictive modeling tools also received important updates in pSeven 6.17:
- More control over model training technique settings in SmartSelection. Previously, you could either run SmartSelection in the default mode, where it selects the best technique automatically, or choose a single specific technique to use, so SmartSelection tunes training settings for this technique only. In pSeven 6.17, you can use the new EnabledTechniques hint to allow two or more certain techniques; SmartSelection then selects the best among them, ignoring the techniques that you disabled.
- The support for importing data fusion models into reports is added. You can now use data fusion models with all predictive modeling tools available in Analyze similarly to approximation models.
- When training models with the ApproxBuilder block, you can add specific input and output bounds. Input bounds define the model’s input domain (valid ranges of inputs); the block can determine them from the training data, or you can adjust the bounds manually. Output bounds add thresholds for model outputs.
- You can now hard stop model training in Analyze.
- You can now copy model output sensitivity scores (Sobol indices) calculated by the Model explorer to the clipboard or export them to a text file.
- Export a dimension reduction model as an Excel document with a linked DLL. This export feature is available through the DRCompressor and DRDecompressor blocks: load a model to the block, then use the export command from the Model file pane. Note that export to Excel is supported only in the Windows version of pSeven.
Data Analysis Updates
- The Sample viewer can now sort the table on the Data tab by any column in the increasing or decreasing value order, similarly to spreadsheet editors.
- Reworked the image export dialogs in tools that support plot export. Added options to adjust image size, select file format, and others.
pSeven 6.17 also includes numerous smaller improvements and bugfixes – please see the release changelog for a full list. You can also contact us to get more information and pSeven updates.