willforan / 02750al Goto Github PK

matlab/run.sh
-- will run on PPI and PF once created (5 times 5 fold cross validation)

plasDB/formatData.sh 
     --> labeled.csv   (with Gene IDs)
     --> labeled.arff  (without Gene IDS)

    use llinas_maxOverAll.txt with > 9 as positive for diff expressed
       - 2 fold change, 70 percentile

matlab/generateMAT.m
    plasDB/labeled.csv --> matlab/genes.mat 
	structure
    	genes.{labels,feats,featnames,names}

matlab/runAL.m
    runAL({'RFRandom' ....},'PF',i)
    * last two inputs optional

matlab/plotAL.m
    plotAL('PF',{'RFRandom' ....})
    * last input optional
   

wekaShortcuts.sh
   source-ing provies

       wekacli:  run weka with arff generated by format data with given classifier and params
          eg. wekacli tree weka.classifiers.trees.J48 -C 0.25 -M 2 

	  saves wekaout/tree-date

       wekainfo: shows % accuracy for a given weka output (likely in wekaout)
         eg. wekainfo wekaout/tree-2011-11-19
       
wekaRuns.sh
    run different learning algorithms (tree, nb, kmeans) [svm is sooo slow]