#TODO: support concatenating multiple datasets
print 'script launched'
try:
from ia3n.util.mem import MemoryMonitor
mem = MemoryMonitor()
except ImportError:
mem = None
if mem:
print 'memory usage on launch: '+str(mem.usage())
import numpy as np
from optparse import OptionParser
from galatea.s3c.hacky_multiclass_logistic import HackyMulticlassLogistic
from galatea.s3c.feature_loading import get_features
from pylearn2.utils import serial
from pylearn2.datasets.cifar10 import CIFAR10
from pylearn2.datasets.cifar100 import CIFAR100
import gc
gc.collect()
if mem:
print 'memory usage after imports'+str(mem.usage())
def train_model(fold_train_X, fold_train_y, C):
fold_train_X = np.cast['float64'](fold_train_X)
model = HackyMulticlassLogistic(C).fit(fold_train_X, fold_train_y)
gc.collect()
return model
def get_labels_and_fold_indices(cifar10, cifar100, stl10):
#TODO: support concatenating multiple datasets
try:
from ia3n.util.mem import MemoryMonitor
mem = MemoryMonitor()
except ImportError:
mem = None
if mem:
print 'memory usage on launch: '+str(mem.usage())
import numpy as np
import warnings
from optparse import OptionParser
try:
from sklearn.svm import LinearSVC, SVC
except ImportError:
from scikits.learn.svm import LinearSVC, SVC
from pylearn2.datasets.tl_challenge import TL_Challenge
from galatea.s3c.feature_loading import get_features
from pylearn2.utils import serial
import gc
gc.collect()
if mem:
print 'memory usage after imports'+str(mem.usage())
def get_svm_type(C, one_against_many):
if one_against_many:
svm_type = LinearSVC(C=C)
else:
svm_type = SVC(kernel='linear',C=C)
return svm_type
#TODO: support concatenating multiple datasets
try:
from ia3n.util.mem import MemoryMonitor
mem = MemoryMonitor()
except ImportError:
mem = None
import numpy as np
import warnings
from optparse import OptionParser
try:
from sklearn.svm import LinearSVC, SVC
except ImportError:
from scikits.learn.svm import LinearSVC, SVC
from galatea.s3c.feature_loading import get_features
from pylearn2.utils import serial
from pylearn2.datasets.cifar10 import CIFAR10
import gc
rng = np.random.RandomState([1,2,3])
def get_svm_type(C):
svm_type = LinearSVC(C=C)
return svm_type
def subtrain(fold_train_X, fold_train_y, C):
assert str(fold_train_X.dtype) == 'float32'
#assert fold_train_X.flags.c_contiguous