Release Notes Version 7 5 1

From Eigenvector Research Documentation Wiki
Revision as of 17:16, 29 January 2014 by imported>Scott (→‎Changes and Bug Fixes in Version 7.5.2)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Changes and Bug Fixes in Version 7.5.2

These changes were primarily for the Solo and Solo_Predictor line of products.

File Comment
Plot Controls
  • Fix for Select Class menu option and toolbar button throwing error
Solo Scripting
  • Fix for indexing error when using ":" shortcut with no starting value
exportfigure
  • Fix for screencapture to clipboard on Mac platform.

Changes and Bug Fixes in Version 7.5.1

File Comment
analysis
  • Fix for LV's not getting updated after crossterm
  • Make sure preprocessing is a ROW vector before adding clutter (avoids indexing error when loading manually-created preprocessing)
  • Fix bug where filternwayprepro option wasn't being saved
  • Add 'whittaker' 'baseline' to methods that are allowed with Multiway data (although they ONLY work on mode 2 of the data)
  • Add exception to allow ncomp > max size data for ANN analyses (because it is # of nodes which is allowed to be > # of inputs)
  • Fix for autorebuild function not existing when options are changed in SIMCA
  • Add "reduced stats" to analysis to make it easier to access reduced statistics in scores plot and make "both" option the default
  • Enable cross-validation for ANN
  • Add GUI-based tests for multivariate Y for methods that can't handle it (when found, send user directly to "select y column" interface)
  • Allow 0 for 2nd layer nodes in ANN settings (meaning do not use a second layer - bug in 7.5 ALWAYS used two layers)
  • Fix for bug in # of components validation in ANN settings (was updating the wrong edit box)
  • Fix for misassigned compression components options field
  • Show PARAFC constraints option at "novice" level of options
  • Fix fontsize on panels to be more reasonable
  • Add help button to mcr constraints gui
  • Include constraints in the items shown with PARAFAC model information
  • Fix for copying figures on MAC
  • Fixed a semi-severe bug in PARAFAC split half analysis which the percentage goodness was in error (The plots though were all good)
ann
  • Add new back-propogation algorithm (faster and more accurate.)
  • Enable cross-validation with full support for all leave-out methods.
  • Add I/O version where number of nodes in one or both layers can be entered as an argument
autoimport
  • Fix for possible explode-into-workspace problem if object is a cell
  • When XLS files are being read from a graphical interface, use graphical_selection as default parsing
boxplot
  • Do not sort classes by name - keep in same order as what was passed in (so user can change order if they want)
  • Fix for bugs associated with incorrect length of input cls and/or with numeric input cls when columns being combined
browse
  • Fix bug for 2014a, clean up combo box code, fix combo box code that was getting called on window button motion. Add test before update.
bspcgui
  • fix for errors in options settings (wrong field name and misspelled function name)
coadd
  • Enable coadd settings GUI to work with multiway data
  • Enable ALL transform methods for multiway in dataset editor
  • Add better error trapping in polytransform to explain the method does NOT work on multiway data
crossval
  • New waitbar logic to improve frequency and accuracy of display (and opportunity to cancel.)
  • Fix legend entries for plots created for multivariate y
dataset
  • Search: Fix for error thrown when vector accidentally given as comparison value (e.g. when user isn't done typing)
  • Search: Allow ~=, <> and = as valid comparison operators
  • Search: Fix for inverted logic when user uses comma (should be considered "or" not "and")
  • Search: Fix for errors thrown when one segment returns empty (and -> always return empty; or -> ignore empty)
evrimodel
  • Fix for calculating cross-validation on more components than asked for (max of # asked for vs. smaller of # var/samples)
  • Consider various "virtual" fields as "true" for isfield tests
  • Add special handling of ncomp for knn method
  • Update ANN models to include cross-validation results fields
Hierarchical Model Builder
  • Fix for dragging of model into end-node.
  • Add "Clear Canvas" option.
  • Add support for testing T2 and Q on rule nodes based on regression and projection models.
Model Optimizer
  • Add "Survey Preprocessing" button, improve performance, fix duplicate checking.
evriscript
  • Add ANN evriscript_module
importtool
  • fix for error thrown when close button is used to close interface (instead of cancel)
  • Fix offset of predefined label columns.
  • Remove unused menu and toolbar.
modelcache
  • Allow export of object and linked data into a single file.
  • Don't pester user about purging old items as much.
plsda
  • If user has passed only class 0 and one other class, do NOT consider class 0 "unknown"
textreadr
  • Add flag for graphical selection of XLS files.
  • Add xlsb to file formats recognized as excel
  • Trust file extension explicitly (remove support for TEXT files with excel extensions; fixes bugs on systems without excel)
  • Make default selection "graphical_selection"
  • Add option to control default parsing selection
  • Better handle case when NO delimiters could be found
  • Fix for error when no numerical values are present OR numerical and text data already match EXACTLY
  • Change logic to assure XLS file text/data alignment is done whenever the matrices don't match (catches some odd shape cases); improve some fringe-case handling of alignment issues
  • Improve memory and speed performance when calling importtool
  • Improve logic to identify label columns
  • Add maxpreviewcols and maxpreviewrows as options to control how many rows/columns are shown in import tool preview
trendtool
  • Instill use of figuretheme
  • Assure trend gets updated when DSO is updated
  • Use stored line styles when dragging markers
  • Various bug fixes
xlsreadr
  • Better handle PCs without excel and MACs (don't call various functions that we know won't work)