Updates 2016/09/11
The replacement of the default Weka interfaces (Explorer, Experimenter) with more user-friendly and efficient interfaces is more or less finished, with the introduction of the Multi-Experimenter (see below).
Fixes
adams.core.discovery.DefaultPropertyDiscovery now uses the regular expressions on the property paths.
adams-twitter: The extraction of hashtags from tweet archives incorrectly cut off the last character from tags (Retwee instead of Retweet).
adams-rats: Fixed the incomplete full names of Rat actors.
Changes
The Find next menu item has been removed from the Flow Editor's menu, as the updated Find menu item now lists all occurrences in a dialog.
The File commander now offers action plugins, with the adams-compress module offering GZIP and ZIP actions.
The RegExp Test maintenance tool now offers Replace and Match tabs.
The FileSystemSearch source is now stoppable.
adams-net: upgraded jsch to 0.1.54.
-
adams-weka: The Weka Investigator received a few changes:
Supports now leave-one-out cross-validation (LOO-CV) and multi-threaded evaluation in the Classify tab.
Sports the 4-in-1 plot available from the adams-visualstats module as output for the classify tab.
In the data tab, you can now display histograms and simple plots for rows/columns.
Undo is now available as well - per dataset level.
Modules have been rearranged: adams-base-all, adams-addons-all annd adams-incubator-all have been moved to adams-applications. The Snapshot (Apps) page on the ADAMS website got removed, as the application snapshots have been folded into the regular snapshots.
Additions
Added object reader/write using a nested command-line format, which is easier to edit manually than the default command-line format.
Added FindText flow processor that allows searching for text in the actors' options (String and BaseObject-derived ones).
adams-net: additional searchlets for the FileSystemSearch source that allow searching on remote servers using SFTP channels. A SSHConnection standalone actor must be present in the flow for this to work.
adams-rats: added Exec rat input and output schemes which execute an external program.
-
adams-weka:
Added Attribute selection panel to the Weka Investigator.
Added Weka Multi-Experimenter tool, which offers a user interface that utilizes the Weka experiment framework mainly for analyzing experimental results. Weka's simple setup is available, as well as one for ADAMS-based experiments. The advantage of the latter is support for multi-threading and (in theory) also distributed computation.
There is a new snapshot available: adams-ml-app. This collection of modules combines all machine learning modules.