def dummy():
## a very simple test
xDummy = numpy.array([[0, 0], [1, 1], [1, 2]])
assert xDummy.shape == (3, 2)
## quadratic basis function
M = 2
phiDummy = makePhi(xDummy, M)
assert phiDummy.shape == (3, 6)
n, m = phiDummy.shape
yDummy = numpy.array([[-1], [1], [1]])
assert yDummy.shape == (3, 1)
# Carry out training, primal and/or dual
C = 10e10
C = 0.25
p = svm.primal(phiDummy, yDummy, C)
w = numpy.array(p['x'][:m])
assert w.shape == (6, 1)
b = p['x'][m]
assert isinstance(b, (int, float))
dummyPredictor = makePredictor(w, b, M, 'svm')
plotDecisionBoundary(xDummy,
yDummy,
dummyPredictor, [-1, 0, 1],
title='SVM Validate')
def _train(self, X, Y):
## save for use in _getPredictor below
self.tX = X
self.tY = Y
## first, cvxopt output suppression
if not self.printInfo: cvxopt.solvers.options['show_progress'] = False
phi = makePhi(X,self.M)
n,m = phi.shape
primal = self.params['primal']
C = self.params['C']
if primal:
self.result = svm.primal(phi,Y,C)
w = numpy.array(self.result['x'][:m])
b = self.result['x'][m]
else:
self.kernel = self.params['kernel']
self.result = svm.dual(phi,Y,C,self.kernel)
self.alphaD = numpy.array(self.result['x'])
w,b,self.dualS,self.dualM = svm.dualWeights(phi, Y, self.kernel, self.alphaD, C)
self.sv = self.numSupport(self.alphaD)
self.slack = numpy.array(self.result['z'])[:-n]
return w,b
def dummy():
## a very simple test
xDummy = numpy.array([[0,0],[1,1],[1,2]])
assert xDummy.shape == (3,2)
## quadratic basis function
M=2
phiDummy = makePhi(xDummy,M)
assert phiDummy.shape == (3,6)
n,m = phiDummy.shape
yDummy = numpy.array([[-1], [1], [1]])
assert yDummy.shape == (3,1)
# Carry out training, primal and/or dual
C = 10e10
C = 0.25
p = svm.primal(phiDummy,yDummy,C)
w = numpy.array(p['x'][:m])
assert w.shape == (6,1)
b = p['x'][m]
assert isinstance(b, (int,float))
dummyPredictor = makePredictor(w,b,M,'svm')
plotDecisionBoundary(xDummy, yDummy, dummyPredictor, [-1, 0, 1], title = 'SVM Validate')
def _train(self, X, Y):
## save for use in _getPredictor below
self.tX = X
self.tY = Y
## first, cvxopt output suppression
if not self.printInfo: cvxopt.solvers.options['show_progress'] = False
phi = makePhi(X, self.M)
n, m = phi.shape
primal = self.params['primal']
C = self.params['C']
if primal:
self.result = svm.primal(phi, Y, C)
w = numpy.array(self.result['x'][:m])
b = self.result['x'][m]
else:
self.kernel = self.params['kernel']
self.result = svm.dual(phi, Y, C, self.kernel)
self.alphaD = numpy.array(self.result['x'])
w, b, self.dualS, self.dualM = svm.dualWeights(
phi, Y, self.kernel, self.alphaD, C)
self.sv = self.numSupport(self.alphaD)
self.slack = numpy.array(self.result['z'])[:-n]
return w, b