GitLab

- Stimulations should now be sent to AS after rendering by the box
- Wrote the missing box documentation
- Minor cleanup

- Added a bash script to loop over boxes and list those that need an update
- Designer will now catch file open/load errors on main level
- Fixed some trivial compiler warnings

- Filename could not be chosen when saving channel names in Acquisition Server, fixed
- Linux gtec driver was not saving/restoring the header, fixed
- Fixed boost chrono find script for Windows, minor cleanup
- Removed unnecessary dependencies from the signal processing plugin
- Fixed VS2013 compiler error in the AS drift correction

- Also sets the current win build target as at least Vista

- Changed the scenarios to use TCP Tagging (software tagging)
- The recorded files will now have only one stimulation stream
- Simplified the scenarios, especially the classification part
- Added lua scripts to filter targets and flashes from P300 timelines
- Added a lua script to accumulate P300 row/column votes
- Added some comments to the scenarios
- Added a replay scenario to the basic P300 speller

- Updated documentation
- Small clean up

- Refactored to use the most recent codec conventions
- Changed to use ITimeArithmetics for time conversions

+ Enabled logging in some Ogre/CEGUI using parts
+ Moved some dox files around, removed one obsolete scenario

- Added a small benchmarking test to study this

- Core components now compile
- Boost 1.55, ITPP, VRPN and Ogre should be provided as VS2013 versions
  and re-enabled

- Refactored the scenarios for simplicity
- Added a .lua evidence accumulator
- Added signal monitoring scenario
- The scenarios are now started and stopped automatically

- Can now optionally send certain stimuli to AS's TCP Tagging when rendering
- Added Boost System dependency for the TCP connection
- Option to only show positive feedback (that is, only to the expected direction)

- Refactored OneVsOne to use map storage
- Pairwise strategies no longer use their own class
count as that can be inherited from base class

n.b. errno state is never cleared by the system functions themselves.
If the state is not manually cleared its value may be unrelated
to the current scope and be a result of some previous call elsewhere.

- Uses the class count from the outside
- Input classes should now be 0,...,k-1
- Allows some classes to have no training examples

- Optional parameter to balance datasets with inequal numbers of labels
- The trainer now provides the classifier with the total number of classes
- The number of classes setting is not shown in the GUI, by design
- Convention: From now on, the classes should be integers from 0,k-1 for k classes.
- Misc code cleanup