def testmmcml(trainx, testx, trainy, testy):
mlmnn = MLMNN.load(MODEL_NAME) # randomstate=32
traind = mlmnn.get_DMatrix(trainx)
from MMCML import MMCML
mmcml = MMCML()
As = [mmcml.fit(traind[:i], trainy, max_iter=50, lr=5e-5).A.flatten() for i in range(1, 11) ]
return As
def testlmnn(trainx, testx, trainy, testy):
mlmnn = MLMNN.load(MODEL_NAME) # randomstate=32
traind = mlmnn.get_DMatrix(trainx)
from LMNN_alpha import LMNN_alpha
lmnn = LMNN_alpha()
As = [lmnn.fit(traind[:i], trainy, max_iter=50, lr=5e-5).A.flatten() for i in range(1, 11) ]
return As
def svmtest(trainx, testx, trainy, testy):
mlmnn = MLMNN.load(MODEL_NAME)
traind = mlmnn.get_DMatrix(trainx, trainx).sum(0)
testd = mlmnn.get_DMatrix(testx, trainx).sum(0)
from sklearn import svm
clf = svm.SVC(kernel='precomputed', C=1.0)
clf.fit(traind, trainy)
print clf.predict(traind)
print trainy
print (clf.predict(traind)==trainy).sum()
print clf.predict(testd)
print testy
print (clf.predict(testd)==testy).sum()
def loadtest(trainx, testx, trainy, testy, As=None):
mlmnn = MLMNN.load(MODEL_NAME)
acc = []
for Time2 in np.linspace(0.1, 1, 10):
Time2 = int(Time2*10)
f = lambda x: x.reshape(x.shape[0], 10, -1)[:, :Time2, :].reshape(x.shape[0], -1)
ttrainx = f(trainx)
ttestx = f(testx)
if As:
acc.append(mlmnn.fittest(ttrainx, ttestx, trainy, testy, Time2, alpha=As[Time2-1][:Time2]))
else:
acc.append(mlmnn.fittest(ttrainx, ttestx, trainy, testy, Time2))
print acc
acc = np.array(acc)
print acc[:,0]
print acc[:,1]
return acc
def loadtest(trainx, testx, trainy, testy, model=None):
if model:
mlmnn = model
else:
print "load temp.MLMNN"
mlmnn = MLMNN.load("temp.MLMNN")
acc = []
for Time2 in np.linspace(0.1, 1, 10):
f = lambda x: x.reshape(x.shape[0], 10, -1)[:, : int(Time2 * 10), :].reshape(x.shape[0], -1)
ttrainx = f(trainx)
ttestx = f(testx)
acc.append(mlmnn.fittest(ttrainx, ttestx, trainy, testy, int(Time2 * 10)))
print acc
acc = np.array(acc)
print acc[:, 0]
print acc[:, 1]
return acc
train_ind, test_ind = train_test_split(range(n), range(n), test_size=0.32, random_state=32)[:2]
# assert ((trainx==x[train_ind]).all())
# assert ((testx==x[test_ind]).all())
duration_model = LMNN_duration.load("./BIT/models/[K=300][rs=32]duration.model")
duration_trainx, duration_trainy = expand(trainx, trainy, K, G)
duration_testx, duration_testy = expand(testx, testy, K, G)
duration_trainx = duration_model.transform(duration_trainx)
duration_testx = duration_model.transform(duration_testx)
duration_knn = KNN(n_neighbor).fit(duration_trainx, duration_trainy)
print duration_trainx.shape
print duration_trainy.shape
prediction = duration_knn.predict(duration_testx)
print prediction.tolist()
mlmnn = MLMNN.load("./BIT/models/[K=300][rs=32]mlmnn.model")
mlmnn_knns = []
trunc = lambda x: x.reshape(x.shape[0], G, -1)[:, :g, :].reshape(x.shape[0], -1)
for g in range(1, G + 1):
mlmnn_knn = KNN(mlmnn.K).fit(mlmnn.transform(trunc(trainx), g), trainy)
mlmnn_knns.append(mlmnn_knn)
for g in range(1, G + 1):
# print '='*10
# acc = 0.
# for count, ind in enumerate(test_ind):
# ttest = mlmnn.transform(trunc(x[ind:ind+1]), g)
# if mlmnn_knns[g-1].predict(ttest)[0]==y[ind]:
# acc += 1
# print acc/(count+1)
import matplotlib
matplotlib.use('Agg')
from tSNE.draw import visualize
import cPickle
if __name__ == '__main__':
K = 100
Time = 1.0
M = 10
trainx, testx, trainy, testy = cPickle.load(open('./features/[K={}][T={}]BoWInGroup.pkl'.format(K, Time),'r'))
print trainx.shape
print trainy.shape
visualize(trainx, trainy, './tSNE/pre-Train.png')
visualize(testx, testy, './tSNE/pre-Test.png')
from MLMNN import MLMNN
mlmnn = MLMNN.load('./temp.MLMNN')
trainx = mlmnn.transform(trainx)
testx = mlmnn.transform(testx)
print trainx.shape
visualize(trainx, trainy, './tSNE/post-Train.png')
visualize(testx, testy, './tSNE/post-Test.png')
x = x.reshape(x.shape[0], -1)
print x.shape
# trainx, testx, trainy, testy = train_test_split(x, y, test_size=1./60.)
# trainx, testx, trainy, testy = cPickle.load(open('{}/[K={}][T={}]BoWInGroup.pkl'.format(featureDir, K, Time),'r'))
# print trainx.shape
# print trainy.shape
train_acc, test_acc = [], []
loo = LeaveOneOut(x.shape[0])
for train_index, test_index in loo:
trainx, testx = x[train_index], x[test_index]
trainy, testy = y[train_index], y[test_index]
mlmnn = MLMNN(granularity=int(Time*G),
K=len(set(y)),
mu=0.5, lmdb=0.25, gamma=5.00,
dim=110, alpha_based=0.0,
normalizeFunction=normalize)
mlmnn.fit(trainx, trainy, testx, testy,
tripleCount=10000,
learning_rate=0.5, alpha_learning_rate=1e-5,
max_iter=2, reset_iter=5, epochs=15,
verbose=True)
acc = loadtest(trainx, testx, trainy, testy, mlmnn)
sys.stdout.flush()
train_acc.append(acc[:, 0])
test_acc.append(acc[:, 1])
train_acc = np.array(train_acc)
test_acc = np.array(test_acc)
