Comparison between support vector algorithm and algebraic perceptron

Introduces the idea of applying the perceptron learning algorithm to high-dimensional linear vector spaces with a scalar product. A linear separation is sought in the high-dimensional space that corresponds to a polynomial separation in the low-dimensional input space. This is similar to the polynomial support vector machines (SVMs) but in contrast to those a non-optimal solution will be found in general. A comparison with SVMs is done with binary images as training data