change to name=value
multi class
dscretizer
change reading so it knows the @meta tag
try wthe which test
seperaete out scoring from the rest
reducing rows columns. does as well as full set?


apps
---------------------------
applications of induction to model-based reasoning: more than just "prediction"

quick baseline for other methods
incremental learning
- theory revision
feature selection
- variance reduction
instance selection
- variance reduction
- optimization
classification
optimization
goal-sensitive learning
- adjust the scoring function
   - effort-minimization (pd vs effort)
contrast set learning
- planning/ repair(constrast lower to higher)
   - not n-step planning (just 1 step)
      - for n-step, use this as the eval function on a beam search
- diagnosis (contast higher to lower)
anomaly detection
- verification (e.g. proof carrying code)
explanation
- from naive bayes
- from clustering
mode indentification (anomaly detection when all prior classes are "known")
monitoring (classifying with weighted classes + anomaly detection)
prediction (after classification, have norms for the local site)
decision support
- argument support
distributed learning
- realtime, anytime
- fusion
- privacy and security