def checkClfPerformanceByCV(): data, testData, fieldMaps, sampleWeights, testSampleWeights = readData(outputDir=rootdir) random_state = 0 trans = Pipeline([('filler', MissingValueFiller()), ('normer', Normalizer())]) X = trans.fit_transform(data.X, data.Y) y = data.Y verb = 1 gbjjn_jobs = 10 maxNumIt = 300 clfs = [ GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.75, learningRate=0.5, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.5, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.75, learningRate=0.05, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.05, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.01, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.01, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.07, numDeemedStuck=15, n_jobs=gbjjn_jobs)] clfs = clfs[2:] res = [cvScores(clf, X, y, n_jobs=24, scoreFuncsToUse='accuracy_score', numCVs=10, doesPrint=False) for clf in clfs] # pool = MyPool(processes=10, initializer=init, initargs=(X, y)) # temp = pool.map_async(cvScoreInnerLoop, clfs) # temp.wait() # res = temp.get() print res for i in range(len(clfs)): print '=' * 10 print clfs[i] print res[i]
pool.close() pool.join() print(result) def func1(**kwargs): print 'in func 1' func2(**kwargs) def func2(a): print 'in func 2' print 'a =', a if __name__ == '__main__': data, testData, fieldMaps, _, _ = readData(outputDir=rootdir) random_states = range(10) useJJ = True # make pipe and allParamsDict # name = 'svc' # step1 = fillertoTry # step2 = normalizerToTry # step3 = (name,classifiersToTry[name]) # pipe, allParamsDict = makePipe([step1, step2, step3]) # pprint(allParamsDict) # print '>'*20 # p = loadObject("H:/allparamsdict") # pprint(getParamsFromIndices([2, 2, 1, 0, 2], p))
numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.05, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.01, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.1, numDeemedStuck=15, n_jobs=gbjjn_jobs), GradientBoost_JJ(learners=[svc_f, svc_m, rf_f, rf_m,gb], verbosity=verb, subsample=0.5, learningRate=0.01, numIterations=maxNumIt, lossFunction=LeastSquaresError(1), stoppingError=0.07, numDeemedStuck=15, n_jobs=gbjjn_jobs)] clfs = clfs[2:] res = [cvScores(clf, X, y, n_jobs=24, scoreFuncsToUse='accuracy_score', numCVs=10, doesPrint=False) for clf in clfs] # pool = MyPool(processes=10, initializer=init, initargs=(X, y)) # temp = pool.map_async(cvScoreInnerLoop, clfs) # temp.wait() # res = temp.get() print res for i in range(len(clfs)): print '=' * 10 print clfs[i] print res[i] if __name__ == "__main__": checkClfPerformanceByCV() data, testData, fieldMaps, sampleWeights, testSampleWeights = readData(outputDir=rootdir) # random_state = 0 # # buildModel(data, testData, fieldMaps, selectedClfs=['JJGradientBoostsimple'], n_jobs=4, writeResults=True, # colNames='all', cvNumSplits=10, random_state=random_state) print '\n>>>>>>>>>>> Done <<<<<<<<<<<<<<'