Release Notes Version 9 0

Changes and Bug Fixes in Version 9.0

Version 9.0 of PLS_Toolbox and Solo is scheduled for released in October, 2021.

Solo is now built with version 2020b of Matlab. This will add major improvements to graphics over previous versions of Solo. Expect to see better performance across many tools and better utilization of modern hardware. Matlab 2020b also enables Python (see below).
Python Integration
- This release introduces several Python methods. In order to use these please follow the instructions to get started: Python Configuration. These steps are necessary to use the new methods. Once configured, try the following:
  - ANNDL - Artificial Neural Network Deep Learning.
  - ANNDLDA - Artificial Neural Network Deep Learning for classification.
  - UMAP - Uniform Manifold Approximation and Projection (Unsupervised).
  - TSNE - t-distributed Stochastic Neighbor Embedding.
  - For more info about PLS_Toolbox Python integration see the wiki Python.
PLOTGUI - Create an axisscale from selected points in a plot of X-block data via context (right-click) menu. This can be used to create a Y-block using create Y from X-block Axis Scale menu item.
KNN
- Select Class Groups interface now available in the KNN Analysis window.
- Add option to use compression.
SIMCA
- Sub models can now use independent preprocessing and included variables from the Analysis interface.
- Building SIMCA model from command line can now pass cell array of individual PCA models (built from the same dataset).

File	Comment
analysis	Can now create Y-block from X-block column, axisscale, or class set. And where appropriate, can choose to delete or exclude selection from X-block.
constrainfit	Add 'exponential' to type of constraints available.
experimentreadr	When splitting data into Cal/Val can now keep replicates based on class set from X or Y block. Also can choose to use Mahalanobis distance or Euclidean distance.
hjyreadr	Horiba Raman (.l6s, .l6m) file importer is now faster and can import larger files (up to a maximum of 2048x2048x1024 elements).
CLS-based residuals	Residuals based on a Classical Least Squares model of the data for use in the development of GLSW and EPO filters.