def trainAndEval(trainInData, trainOutData, trainWeights, testInData,
testOutData, optionList, lossFun):
optionCount = len(optionList)
loss = np.zeros((optionCount, ))
trainer = jrboost.BoostTrainer(trainInData, trainOutData, trainWeights)
loss = trainer.trainAndEval(testInData, testOutData, optionList, lossFun)
return loss
def trainAndPredict(trainInData, trainOutData, validationInData, bestOptList):
predOutDataList = []
trainer = jrboost.BoostTrainer(trainInData, trainOutData)
for opt in bestOptList:
predictor = trainer.train(opt)
predOutDataList.append(predictor.predict(validationInData))
predOutData = np.median(np.array(predOutDataList), axis=0)
return predOutData
def trainAndEval(inData, outData, samples, foldCount, optionList, lossFun):
optionCount = len(optionList)
loss = np.zeros((optionCount, ))
folds = jrboost.stratifiedRandomFolds(outData, foldCount, samples)
for trainSamples, testSamples in folds:
trainInData = inData[trainSamples, :]
testInData = inData[testSamples, :]
trainOutData = outData[trainSamples]
testOutData = outData[testSamples]
trainer = jrboost.BoostTrainer(trainInData, trainOutData)
loss += trainer.trainAndEval(testInData, testOutData, optionList,
lossFun)
return loss
def trainAndPredict(inData,
outData,
trainSamples,
testSamples,
opt,
rankFun=None):
trainInData = inData[trainSamples, :]
testInData = inData[testSamples, :]
trainOutData = outData[trainSamples]
trainer = jrboost.BoostTrainer(trainInData, trainOutData)
predOutDataList = []
for opt1 in opt:
predictor = trainer.train(opt1)
predOutData = predictor.predict(testInData)
predOutDataList.append(predOutData)
predOutData = np.median(np.array(predOutDataList), axis=0)
return predOutData