Learning with Permutably Homogeneous Multiple-Valued Multiple-Threshold Perceptrons