Only One Neuron either N-bit Parity Rule Based Modified Translated Multiplicative or McCulloch-Pitts Models for Some Machine Learning Problems

Abstract

In this study, solutions to machine learning problems such as Monk’s 2 (M2), Balloon and Tic-Tac-Toe problems employing a single neuron dependent on rules which use either modified translated multiplicative (πm) neuron or McCulloch-Pitts neuron model is proposed. Since M2 problem is similar to N-bit parity problem, translated multiplicative (πt) neuron model is modified for M2 problem. Also, McCulloch-Pitts neuron model is used to increase classification performance. Then either πm or McCulloch-Pitts neuron model is applied to Balloon and Tic-Tac-Toe problems. When the result of proposed only one πm neuron model that is not required any training stage and hidden layer is compared with the other approaches, it shows satisfactory performance.