"svr") sigma = [0] testimg, testdot, testmapping, testtags = Counter.prepareData(img, dot, sigma, normalize = False, smooth = True ) #print "blub", testimg.shape #print testimg #print testdot, np.sum(testdot) boxConstraints = [] #boxConstraints = [(12, img[:,:,:])] #boxConstraints = [(3, img[0:30,0:30,:])] #boxConstraints.reshape((-1, boxConstraints.shape[-1])) success = Counter.fitPrepared(testimg[testmapping,:], testdot[testmapping], testtags, epsilon = 0.000, boxConstraints = boxConstraints) #success = Counter.fitPrepared(testimg[indices,:], testdot[indices], testtags[:len(indices)], epsilon = 0.000) #print Counter.w, Counter. print "learning finished" #conversion step #Q = kernelize(B, method = "gaussian") ##Q = B * B.transpose() #tags = np.zeros(numVariables,dtype=np.int8) #tags[0:len(pindices)] = 1 #tags[len(pindices):] = -1 #c = dot[allIndices] * (-tags)+ epsilon #upperBounds = [None, pMult, lMult] #success,solution = optimize(tags,Q,c,upperBounds) ## Put model data into dense matrices #print Counter.b, Counter.w