Release Notes
PLS_Toolbox version 4.2 includes a number of new tools
in our graphical interface as well as new command-line functionality.
The following is a list of significant changes in Version 4.2 of
PLS_Toolbox.
If you're looking for changes between version
4.0 to version 4.1, see them here.
Topics
In addition to the listed changes, Version 4.2 also
includes a number of bug fixes and general usability improvements
which are not listed. If you have specific questions, please direct
them to helpdesk@eigenvector.com
[Back
to top]
Graphical Interface Changes
Analysis / Workspace Browser
- Model cache and view tool. Tracks and caches all data and model
settings during the modeling process. [Show
Me!]
- Auto-Select logic for automatic selection of PCs/factors. [Show
Me!]
- Scores button presents a multi-plot view of the typical plots
to review for a given model type (for customization, see plotscores_defaults
function) [Show
Me!]
- Addition of CLS and KNN into Analysis.
- IPLS, IPCR, Stepwise-MLR variable selection methods available
in the Analysis GUI
- Additional clustering algorithms added: Ward's Method, Furthest
Neighbor, Average Paired Distance, Median, Centroid
- New correlation map variable-reordering options
Preprocessing
- New baseline method (selected points) added.
[Back
to top]
High-Level Function Changes
CLS
- New command line function for performing Classical Least Squares
with
standard PLS_Toolbox model format.
CORRSPEC
- New command line function for performing correlation spectroscopy.
See
"corrspec_tutorial.pdf" in the /help folder.
CLUSTER
- Added five new methods:
o Furthest Neighbor
o Average Paired Distance
o Median
o Centroid
o Ward's Method
COMPAREVARS
- Compares any two objects or variables for equality. If unequal,
returns text
description of difference.
New Function Summary
This release includes the following new command-line functions:
alignpeaks.m - Calibrates wavelength scale using standard peaks.
alignspectra.m - Calibrates wavelength scale using standard spectrum.
cls.m - Classical Least Squares regression for multivariate Y.
corrspec.m - Resolves correlation spectroscopy maps.
dispmat.m - Calculates the dispersion matrix of two spectral data
sets.
modelcache.m - Stores and retrieves models in the model cache.
dems/alignpeaksdemo.m
dems/alignspectrademo.m
dems/corrspecdemo.m
dems/data_mid_IR.mat
dems/data_near_IR.mat
dems/mcrdemo.m
help/modelcachehelp.html
utilities/addhelpyvars.m
utilities/baselineset.m
utilities/cachestruct.m
utilities/choosecomp.m
utilities/cls_guifcn.m
utilities/corrspecengine.m
utilities/corrspecutilities.m
utilities/evritree.m
utilities/iplsgui.fig
utilities/iplsgui.m
utilities/makepredhelp.m
utilities/panelmanager.m
utilities/plot_corr.m
utilities/plotscores_cls.m
utilities/plotscores_defaults.m
utilities/resolve_maps_2d_cor.m
utilities/resolve_spectra_2d_cor.m
Known Issues in Version 4.2
The following are known issues in version 4.2.
-
Bias calculations are made and shown only
for calibration data sets. Bias in prediction datasets is
not reported.
- Model cache does not work with Matlab 7.0.1. Upgrade to 7.0.4
to resolve.
- PLS_Toolbox installer and other code not compatible with Matlab
2008a pre-release.
- Crossval function, when using custom cross-validation vector,
does not correctly exclude samples marked as "never use"
(cvi = 0). Work-around: exclude samples from dataset (or hard-delete
from matrix) prior to calling crossval. NOTE: Only an issue when
calling with CUSTOM CVI input. Does not effect normal behavior.
- HTML version of help documentation is not accessed in Matlab
2006a or later. Use "docsearch" as work-around. For
example:
docsearch simpls
- pcaengine allows calling with DataSet object but does not use
include field to index. Work-around, extract data before calling
PCAENGINE with DataSet object (In the future, pcaengine will return
an error when a DataSet object is passed.)
|
