To Do List

New Features

• prtDataSetBigData
• Easy ways to handle ordinal, string, features
• Weka interface
• prtDataSetImage
• http://research.google.com/pubs/pub40814.html Locally Sensitive Hashing tools

Gripes

• KNN with mahalanobis distance is painfully slow
• KNN with observationNames is slow (is this just KNN?)
• RVM is painfully slow with a moderately large number (2000) observations
• No one knows how to use RVs but Kenny
• Feature selection should include some plotting method (how can we unify for all methods?)
• Not enough feature selection methods
• Feature selection should have a way that all prtFeatSels can only use their top performing set
• When plotting algorithms the nameAbbreviation may not be a unique identifier (optimize() has a method for this)
• RVs allow cat() but methods don't account for it. (Don't allow cat, but instead make them cells)
• addParameter / addOptional issues. Do something about this. subclass input parser?