Piecewise-linear classifiers, formal neurons and separability of the learning sets | IEEE Conference Publication | IEEE Xplore