Updates 2024/07/17

This time round, there have been a bunch of important bug-fixes and additions. The Weka Investigator can now perform repeated cross-validation, calculating statistics across the runs (the Experimenter does this per fold, the Investigator per full CV) and also a new tab for executing an experiment run on the current dataset and the chosen classifier.

Last, but not least, ADAMS turned 15 today (first commit in 2009). :-)

Fixes

  • adams-pdf: PDF fonts now properly parse the attributes in the font string (Arrays.binarySearch on unsorted array returned invalid index for Italic).

  • adams-weka and adams-weka-lts:

    • The Multi-Experimenter now sorts the results according to dataset/classifier order (dataset first).

    • Leave-one-out cross-validation (LOOCV) now works correctly in the Multi-Experimenter.

    • The Rename... action for renaming a column of an InstancesTable now uses the correct column index.

    • Changed the R^2 calculation to https://en.wikipedia.org/wiki/Coefficient_of_determination#Definitions

Changes

  • adams-core:

    • spreadsheet formulas can now use median as well

    • upgraded commons-io to 2.16.1

  • adams-imaging:

    • upgraded opex4j to 0.0.3

    • the ObjectCentersFromReport plugin in the Preview browser now allows changing the report reader

    • object finders now have a quiet flag to suppress warnings about -1 indices

  • adams-spreadsheet:

    • The SpreadSheetAggregate transformer now offers the 25th and 75th quartile as well.

  • adams-weka and adams-weka-lts:

    • upgraded common-csv-weka-package to 2024.7.5

Additions

  • adams-imaging:

    • The image object filter ObjectsInPolygon finds annotations that fall within the specified polygon.

    • The CountObjectsInPolygon transformer outputs the number of annotations that fall within the specified polygon.

  • adams-weka and adams-weka-lts: