Release Notes Version 7 0

From Eigenvector Documentation Wiki
Jump to: navigation, search

Version 7.0 of PLS_Toolbox and Solo was released in October, 2012.

For general product information, see PLS_Toolbox Product Page. For information on Solo, see Solo Product Page. This release was done in conjunction with MIA_Toolbox / Solo+MIA version 2.8

(back to Release Notes PLS Toolbox and Solo)

New Features in Solo and PLS_Toolbox

Batch Statistical Process Control Tools

  • New top-level data processor to read, align, tag, and arrange batch data into appropriate form for batch analysis.
  • Creates data in appropriate format for analysis with these model types:
    • Summary PCA (PCA on summary of variables over time)
    • Batch Maturity (PCA with heterogeneous confidence limits)
    • MPCA (Multiway PCA)
    • PARAFAC (Parallel Factor Analysis)
    • Summary PARAFAC (PARAFAC on summary of variables over time)
    • PARAFAC2 (only available in PLS_Toolbox with MATLAB)
  • Graphical and automatic identification of batches and [optional] steps in the imported data.
  • Automatic alignment of batches (when necessary) by linear, infilling, or Correlation Optimized Warping.
  • Summary methods allow a wide range of statistics to be calculated for each variable.
  • Opens processed data directly in Analysis for immediate model building.
  • Steps to process data stored for easy application to new data (in data application mode.)

Analysis Window

  • BatchMaturity analysis type added (PCA model with heterogeneous confidence limits for scores).
  • Split data into calibration / validation sets using manual or automatic selection. Movie.png View Video
  • Calculate relative T^2 and Q contributions. New buttons on Plot Controls allow selection of sample(s) as a T or Q reference set. Resulting T or Q contributions are done relative to those selected sample(s). Movie.png View Video
  • Y-block loadings included in bi-plots (PLS).
  • Cross-validation results in SSQ table. Movie.png View Video
  • 3D Loadings from multiway methods can be plotted as 3D surfaces (or other 3D plots).
  • Change included data directly on preprocessed data plots.
  • "Export to Regression Vector" allowed for MLR models.
  • Cross-validation default enabled with improved user awareness of options.
  • Model Cache "Date" mode now sorts in descending order (for faster access to the most recent models and data)
  • Tucker congruence and core consistency test added for multiway models (warn user if it looks like the "supposed to be one" components in the core are showing signs of degeneracy.)
  • Purity now has "Resolve" and "Accept" buttons to improve usability.

Scores Plots

(see also #Plot Controls and Visualization Tools below)

  • Double-sided confidence limits display is more configurable: display as shaded regions, lines, or both and choose color.
  • Cross-validation sub-sets are included as classes to show which samples were in which cross-validation groups. Movie.png View Video
  • Add support for SVM One-Class models (command-line only)

Plot Controls and Visualization Tools

  • Automatic find and mark peak locations. Movie.png View Video
  • Quick-search bar for selecting by labels, axisscales, classes and indexes. Movie.png View Video
  • Plot type button for quick change between various plot types. Movie.png View Video
  • Plot types "Monotonic", "scatter" and "line" added.
  • Right-click access to adjust axis limits and other plot settings.
  • Improved appearance of selections, 3D plots, stacked plots, and class set identifiers.
  • Class population statistics by right-clicking data (shows # and % of samples in each class). Movie.png View Video
  • Colormap changes by right-clicking image (With MIA_Toolbox and Solo+MIA only).
  • Autosize makers option added ( automatically adjusts marker sizes to match axes size, if not specified otherwise).
  • Color-by axisscales and index and clarify what will be colored for all types (lines or points).
  • Magnify tool now has an easy "resize" corner.
  • Label points by their axisscale value.

TrendTool

  • Automatically find peaks and place markers at those peaks.
  • Images now allow showing of more than 3 markers at a time (MIA_Toolbox and Solo+MIA only).
  • Image axisscale is used when displaying images (MIA_Toolbox and Solo+MIA only).

DataSet Editor

  • Axistype (continuous, discrete,...) support in labels tabs.
  • Bulk selection changes in label tab context menus (allows quick selections based on list of all samples.)
  • Export to ThermoGalactic SPC file format from File menu.
  • Data drop support (drop onto tabs imports data.)
  • Data augmentation adds classes to identify different data blocks (for column-augment mode.)

Import / Export

  • SPC File Format
    • Improved multiple file reading (with unequally spaced x-axis)
    • improved handling of automatic axis scale names
  • CSV File Format -Allow space, tab, and | as valid automatically-detected delimiters for CSV files (improves drag/drop importing behavior).

Preprocessing and Transformations

  • savgol -New "tails" mode to improve performance at ends of spectra.
  • classcentroid -New class centroid centering preprocessing method.
  • mscorr -MSC with new 'median' method for robust scaling (and to use with Probabilistic Quotient Normalization - PQN.)
  • wlsbaseline -Whittaker filter option added to Weighted Least Squares baseline. FAST and better for baselines which don't look like polynomials.
  • reducennsamples -Added access to help within settings dialog.

New Demo Datasets

  • cancer -Fluorescence EEM spectra from images of cervices with various states of cervical cancer.
  • Dupont_BSPC -Batch data (10 variables x 36 batches at 100 time intervals) from Nomikos & MacGregor, Technometrics, 37(1), 1995.
  • OliveOilData -Olive Oil FT-IR Classification data from Dahlberg, et. al. Appl. Spectrosc., 51(8), 1118-1124 (1997)

General Solo Improvements

  • Re-enable docked figures with Solo & Solo+MIA.
  • Improved memory performance (java.opts modification).


New Command-line Features and Functions

  • Full Support for Matlab R2012b

EVRIModel Objects

New high-level object to contain models. Provides an object-oriented approach to building new models and makes working with models and predictions much easier.

  • Build models using simple object-oriented assignments and methods
  • Apply models to new data directly using model methods: prediction = model.apply(newdata);
  • Review results in plots using object methods: model.plotscores, model.plotloads, model.ploteigen
  • Access commonly-used results via simple properties: .scores, .predictions, .tcon, .t2 instead of having to remember array indexes and cryptic field names.
  • Nearly complete backwards compatibility with existing user code.

Command-line Tool Changes

  • Quick Reference Card -New quick reference card ( PLS_Toolbox_Quick_Reference.pdf )
  • autoexport -add SPC export functionality.
  • chitest -add distribution name and function name to chitest outputs (making it much easier to apply the results).
  • coreanal -updated coreanal.m to be able to provide a list of important core values (new optional second output).
  • crossval -added output of cvi to help identify which leave-out group each sample was in.
  • encode -Increase number of items allowed in each row of "speed" encoded files (makes the encoding MUCH faster)
  • ils_esterror -Various improvements to allow different types of error estimates.
  • mscorr -Add option.algorithm to include new option 'median', based on Probabilistic Quotient Normalization.
  • spcreadr
    • Improved multiple file reading (with unequally spaced x-axis)
    • Improved handling of automatic axis scale names
  • svmoc -add support to plot scores from SVM One Class models.
  • windowfilter -Added method 'roll' (for processing rows only), slight modification to RH edge indexing during call (is last channel processed?)
  • wlsbaseline -Add Whittaker filter option to wlsbaseline and wlsbaselineset (FAST and better for baselines which don't look like polynomials)
  • xclreadr -Allow space, tab, and | as valid automatically-detected delimiters for CSV files
  • DataSet Object - Changes:
    • Decrease dependency on PLS_Toolbox
    • Allow assignment directly onto imageaxisscale
  • DataSet Object - New Methods:
    • FINDSET -Locate a set within a label field (axisscale,label,class) in a DataSet.
    • LISTSETS -For a given field and mode list the sets available.
    • SEARCH -Search for given term in a dso field, mode, and set.
    • UPDATESET -Add/update a label field (axisscale,label,class) in a DataSet.

Misc New Functions

batchalign - Convert data columns based on matching ref col to target vector.
batchmaturity - Batch process model and monitoring.
batchfold - Transform batch data into dataset for analysis.
classcentroid - Centers data to the centroid of all classes.
evrimodel - EVRI Model Object.
minimizemodel - Shrinks model by removing non-critical information.
plotmonotonic - Plot lines with breaks when the x-value "doubles-back" on itself.
roccurve - Calculate and display ROC curve(s) for yknown and ypred.
splitcaltest - Splits randomly ordered data into calibration and test sets.
unhist -Create a vector whose values follow an empirical distribution.
writespc - Writes Galactic SPC files.