def __init__(self):
self.mc = MultiClassifier()
self.labelset = set()
self.c = 1.0
self.j = None
self.tmp = "/tmp/"
self.binary = "svm_perf_learn"
class SVMMultiClassifier:
def __init__(self):
self.mc = MultiClassifier()
self.labelset = set()
self.c = 1.0
self.j = None
self.tmp = "/tmp/"
self.binary = "svm_perf_learn"
def train(self, ds):
bds = ds.binarize()
positions = None
print "SVMMultiClassifier: Training with %d docs, %d labels" % (len(ds.docs), len(bds))
for label in bds:
if positions is None:
fout = tempfile.NamedTemporaryFile(suffix='svm', dir=self.tmp)
positions = bds[label].toSVM(fout)
else:
bds[label].toSVMSubsequent(fout, positions)
fout.file.flush()
modelfilename = "%s-%s.model" % (fout.name, label)
_run("%s %s %s %s > /dev/null 2>&1" % (self.binary, _svm_params(self), fout.name, modelfilename))
self.mc.add(label, SVMClassifier(modelfilename))
os.remove(modelfilename)
fout.close()
self.labelset = set(bds)
def __repr__(self):
return "<SVMMultiClassfier: %d labels>" % len(self.labelset)
def score(self, av):
return self.mc.score(av)
class SVMMultiClassifier:
def __init__(self):
self.mc = MultiClassifier()
self.labelset = set()
self.c = 1.0
self.j = None
self.tmp = "/tmp/"
self.binary = "svm_perf_learn"
def train(self, ds):
bds = ds.binarize()
positions = None
print "SVMMultiClassifier: Training with %d docs, %d labels" % (len(
ds.docs), len(bds))
for label in bds:
if positions is None:
fout = tempfile.NamedTemporaryFile(suffix='svm', dir=self.tmp)
positions = bds[label].toSVM(fout)
else:
bds[label].toSVMSubsequent(fout, positions)
fout.file.flush()
modelfilename = "%s-%s.model" % (fout.name, label)
_run("%s %s %s %s > /dev/null 2>&1" %
(self.binary, _svm_params(self), fout.name, modelfilename))
self.mc.add(label, SVMClassifier(modelfilename))
os.remove(modelfilename)
fout.close()
self.labelset = set(bds)
def __repr__(self):
return "<SVMMultiClassfier: %d labels>" % len(self.labelset)
def score(self, av):
return self.mc.score(av)
def __init__(self):
self.mc = MultiClassifier()
self.labelset = set()
self.c = 1.0
self.j = None
self.tmp = "/tmp/"
self.binary = "svm_perf_learn"
