def constructFromFile(self, fileName, **args):
matrix = ckernel.KernelMatrix()
matrix.thisown = 0
patternID = []
delim = None
delim = misc.getDelim(fileName)
idColumn = 0
if 'idColumn' in args:
idColumn = args['idColumn']
if idColumn is None:
firstColumn = 0
else:
firstColumn = 1
print firstColumn
print idColumn
matrixFile = myio.myopen(fileName)
firstRow = True
for line in matrixFile:
# skip comments:
if line[0] in ["%", "#"]: continue
tokens = line.split(delim)
# check if the file is in gist format:
if firstRow:
firstRow = False
try:
float(tokens[-1])
except:
continue
if (('headerRow' in args and args['headerRow'])
or ('gistFormat' in args and args['gistFormat'])):
continue
values = arrayWrap.floatVector(
[float(token) for token in tokens[firstColumn:]])
matrix.addRow(values)
if idColumn is not None:
patternID.append(tokens[0])
ckerneldata.KernelData.__init__(self, matrix)
if 'labelsFile' in args:
self.attachLabels(labels.Labels(args['labelsFile'], **args))
else:
self.attachLabels(labels.Labels(None, patternID=patternID))
def load_libsvm_format(file_name, **args):
"""
Load a dataset from a file in libsvm format
returns an instance of PyVectorDataSet
If you want to use the data with a SparseDataSet, you can directly
do it using the SparseDataSet constructor.
"""
regression = False
if 'regression' in args:
regression = args['regression']
# first extract labels and check how many features there are:
labels = []
num_features = 0
if not os.path.exists(file_name):
raise ValueError, "file doesn't exist at %s" % file_name
file_handle = myio.myopen(file_name)
for line in file_handle:
tokens = line.split()
if regression:
labels.append(float(tokens[0]))
else:
labels.append(str(int(float(tokens[0]))))
for token in tokens[1:]:
id, value = token.split(':')
num_features = max(num_features, int(id))
X = numpy.zeros((len(labels), num_features), numpy.float)
# fill in the array:
i = 0
for line in open(file_name):
tokens = line.split()
for token in tokens[1:]:
id, value = token.split(':')
id = int(id) - 1
X[i][id] = float(value)
i += 1
data = PyVectorDataSet(X)
if regression:
labels = Labels(labels, numericLabels=True)
else:
labels = Labels(labels)
data.attachLabels(labels)
return data
def load_libsvm_format(file_name, **args) :
"""
Load a dataset from a file in libsvm format
returns an instance of PyVectorDataSet
If you want to use the data with a SparseDataSet, you can directly
do it using the SparseDataSet constructor.
"""
regression = False
if 'regression' in args :
regression = args['regression']
# first extract labels and check how many features there are:
labels = []
num_features = 0
if not os.path.exists(file_name) :
raise ValueError, "file doesn't exist at %s" % file_name
file_handle = myio.myopen(file_name)
for line in file_handle :
tokens = line.split()
if regression :
labels.append(float(tokens[0]))
else :
labels.append(str(int(float(tokens[0]))))
for token in tokens[1:] :
id,value = token.split(':')
num_features = max(num_features, int(id))
X = numpy.zeros((len(labels), num_features), numpy.float)
# fill in the array:
i = 0
for line in open(file_name) :
tokens = line.split()
for token in tokens[1:] :
id,value = token.split(':')
id = int(id) - 1
X[i][id] = float(value)
i+=1
data = PyVectorDataSet(X)
if regression :
labels = Labels(labels, numericLabels=True)
else :
labels = Labels(labels)
data.attachLabels(labels)
return data
def constructFromFile(self, fileName, **args):
matrix = ckernel.KernelMatrix()
matrix.thisown = 0
patternID = []
delim = None
delim = misc.getDelim(fileName)
idColumn = 0
if "idColumn" in args:
idColumn = args["idColumn"]
if idColumn is None:
firstColumn = 0
else:
firstColumn = 1
print firstColumn
print idColumn
matrixFile = myio.myopen(fileName)
firstRow = True
for line in matrixFile:
# skip comments:
if line[0] in ["%", "#"]:
continue
tokens = line.split(delim)
# check if the file is in gist format:
if firstRow:
firstRow = False
try:
float(tokens[-1])
except:
continue
if ("headerRow" in args and args["headerRow"]) or ("gistFormat" in args and args["gistFormat"]):
continue
values = arrayWrap.floatVector([float(token) for token in tokens[firstColumn:]])
matrix.addRow(values)
if idColumn is not None:
patternID.append(tokens[0])
ckerneldata.KernelData.__init__(self, matrix)
if "labelsFile" in args:
self.attachLabels(labels.Labels(args["labelsFile"], **args))
else:
self.attachLabels(labels.Labels(None, patternID=patternID))
def __init__(self, file, **args):
if type(file) == type(''):
if not os.path.exists(file):
raise ValueError, "file does not exist at %s" % file
self._fileHandle = myio.myopen(file)
#self._fileHandle = open(file)
else:
self._fileHandle = file
if 'classes' in args:
self.classesToRead = args['classes']
else:
self.classesToRead = []
if 'patterns' in args:
self.patternsToRead = args['patterns']
else:
self.patternsToRead = None
if 'features' in args:
self.featuresToRead = args['features']
else:
self.featuresToRead = []
def __init__(self, file, **args) :
if type(file) == type('') :
if not os.path.exists(file) :
raise ValueError, "file does not exist at %s" % file
self._fileHandle = myio.myopen(file)
#self._fileHandle = open(file)
else :
self._fileHandle = file
if 'classes' in args :
self.classesToRead = args['classes']
else :
self.classesToRead = []
if 'patterns' in args :
self.patternsToRead = args['patterns']
else :
self.patternsToRead = None
if 'features' in args :
self.featuresToRead = args['features']
else :
self.featuresToRead = []
