Release Notes Version 9 5 0

Changes and Bug Fixes in Version 9.5

Version 9.5 of PLS_Toolbox and Solo is scheduled for released in October, 2024. See our ZZZZ webinar on new features.

(back to Release Notes PLS Toolbox and Solo)

Diviner

Semi-automated modeling that includes preprocessing, variable selection, and outlier removal. Semi-automated means the model creator is kept in the loop so their domain-specific knowledge can be used to guide the analysis.

• Select an Eigenvector provided preprocessing library for your domain (NIR, Raman, discrete,...) or build your own.
• Detect and review outliers using robust methods.
• Generate a first round of model candidates and review performance. Use key plots showing overfit ratio vs cross-validation error (RMSECV) to find high performing models.
• Identify the models with best performance and further tune preprocessing, variable selection, and outlier status.
• Select a final model and have confidence that an optimal set of parameters have been used to create the model.

Ensemble Models

Predictions based on a collection of regression models.

• Build an ensemble from a collections of any number of models.
• Use model fusion or voting regression to generate a prediction.
• See a model's calibration error as the root mean squared error of the aggregated predictions from each of the models.

Importers

• abbspectrumreadr - Reads an ABB SPECTRUM file into a DataSet object (requires python).

Other Changes

File	Comment
browse	Add menu item to adjust default colormap.
modelcache	Update derby database to version 10.14 and update class to new syntax in EVRICACHEDB class.
evrishapley	Add support for Ensemble models.
parafac	Add auto_outlier option to options interface.
preprocess	Save preprofavorites.mat file to EVRI home directory instead of MATLAB prefs folder.
testrobustness	Check Y-block and account for excluded variables.