def recallWithParams(filePrefix, rootPath, tru, tri, teu, topN, featureK,
model):
usersVectorFile = filePrefix + 'usersVector.npy'
usersBiasFile = filePrefix + 'usersBias.npy'
itemsVectorFile = filePrefix + 'itemsVector.npy'
itemsBiasFile = filePrefix + 'itemsBias.npy'
thetaPath = rootPath + 'final.gamma' + str(featureK)
usersVector = utils.loadNumpyMatrix(usersVectorFile)
itemsVector = utils.loadNumpyMatrix(itemsVectorFile)
usersBias = None #utils.loadNumpyMatrix(usersBiasFile)
itemsBias = None #utils.loadNumpyMatrix(itemsBiasFile)
trainUser, trainData = cu.loadData(tru, 'user')
trainItem, noUse = cu.loadData(tri, 'item')
if hasLDA:
theta = cu.loadTheta(thetaPath)
else:
theta = []
userNum = len(trainUser.keys())
itemNum = len(trainItem.keys())
noUse = None
#print 'load test data...'
testUser, testData = cu.loadData(teu, 'user')
#testItem, testDataNoUse = cu.loadData(tei, 'item')
noUse = None
recallAll(trainUser, trainItem, testUser, itemNum, topN, usersVector,
itemsVector, usersBias, itemsBias, featureK, model, theta)
def solve(argv, alpha, Lambdau, Lambdav, Lambdauv, epochs):
print argv
MF = True
CTR = False
RAPMF = True
RAPARE = False
hasLDA = False
save = True
dataset = argv[0]
if dataset == 'delicious' or dataset == 'fm' or dataset == 'blog':
scale = 1
else:
scale = 5
kernel = argv[1]
initFunction = argv[2]
featureK = argv[3]
tru = argv[4]
tri = argv[5]
teu = argv[6]
tei = argv[7]
topN = [1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50]
thetaPath = argv[8]
filePrefix = argv[9]
print 'load train data...'
trainUser, trainData = cu.loadData(tru, 'user')
trainItem, noUse = cu.loadData(tri, 'item')
userNum = len(trainUser.keys())
itemNum = len(trainItem.keys())
noUse = None
print 'load test data...'
testUser, testData = cu.loadData(teu, 'user')
noUse = None
print 'User No.', userNum, 'Item No.', itemNum, 'ratings in Train:', len(
trainData), 'rating in Test:', len(testData)
mean = 0.0
theta = []
if RAPARE or RAPMF:
if RAPARE:
print 'load theta...'
theta = cu.loadTheta(thetaPath)
else:
theta = []
usersVector = None
itemsVector = None
usersBias = None
itemsBias = None
usersRatingCountedSet = utils.countRatings(trainUser)
itemsRatingCountedSet = utils.countRatings(trainItem)
ra.train(mean, userNum, itemNum, featureK, trainData, testData, epochs,
alpha, Lambdau, dataset, usersRatingCountedSet,
itemsRatingCountedSet, True, usersVector, itemsVector,
usersBias, itemsBias, Lambdauv, theta,
filePrefix + str(featureK), save, scale, Lambdav, hasLDA, tru,
tri, teu, topN)
def solve(argv, alpha, Lambdau, Lambdav, Lambdauv, epochs):
print argv
MF = False
CTR = False
RAPMF = False
RAPARE = True
hasLDA = False
save = True
dataset = argv[0]
if dataset == 'delicious' or dataset == 'fm':
scale = 1
else:
scale = 5
newElement = argv[1]
kernel = argv[2]
initFunction = argv[3]
featureK = argv[4]
tru = argv[5]
tri = argv[6]
teu = argv[7]
tei = argv[8]
topN = [1, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50]
thetaPath = argv[9]
filePrefix = argv[10]
print 'feature:', featureK
print 'load train data...'
trainUser, trainData = cu.loadData(tru, 'user')
trainItem, noUse = cu.loadData(tri, 'item')
userNum = len(trainUser.keys())
itemNum = len(trainItem.keys())
noUse = None
print 'load test data...'
testUser, testData = cu.loadData(teu, 'user')
#testItem, testDataNoUse = cu.loadData(tei, 'item')
noUse = None
print 'User No.', userNum, 'Item No.', itemNum, 'ratings in Train:', len(
trainData), 'rating in Test:', len(testData)
#mean = 4.31725404304#video
#mean = 4.20794266709#google play
mean = 4.18519626023 #auto
mean = 0.0
theta = []
if MF:
print 'start matrix factorization'
mf.train(userNum, itemNum, featureK, trainData, testData, epochs,
Lambdau, Lambdav, alpha, mean, filePrefix + str(featureK),
save, theta, scale)
if CTR:
print 'load theta from...', thetaPath
theta = cu.loadTheta(thetaPath)
print 'start CTR'
mf.train(userNum, itemNum, featureK, trainData, testData, epochs,
Lambdau, Lambdav, alpha, mean, filePrefix + str(featureK),
save, theta, scale)
if RAPARE or RAPMF:
if RAPARE:
print 'load theta...'
theta = cu.loadTheta(thetaPath)
else:
theta = []
usersVector = None
itemsVector = None
usersBias = None
itemsBias = None
usersRatingCountedSet = utils.countRatings(trainUser)
itemsRatingCountedSet = utils.countRatings(trainItem)
ra.train(mean, userNum, itemNum, featureK, trainData, testData, epochs,
alpha, Lambdau, dataset, usersRatingCountedSet,
itemsRatingCountedSet, True, usersVector, itemsVector,
usersBias, itemsBias, Lambdauv, theta,
filePrefix + str(featureK), save, scale, Lambdav, hasLDA, tru,
tri, teu, topN)