Maximum 2-satisfiability in radial basis function neural network

Abstract

Maximum k-Satisfiability (MAX-kSAT) is the logic to determine the maximum number of satisfied clauses. Correctly, this logic plays a prominent role in numerous applications as a combinatorial optimization logic. MAX2SAT is a case of MAX-kSAT and is written in Conjunctive Normal Form (CNF) with two variables in each clause. This paper presents a new paradigm in using MAX2SAT by implementing in Radial Basis Function Neural Network (RBFNN). Hence, we restrict the analysis to MAX2SAT clauses. We utilize Dev C++ as the platform of training and testing our proposed algorithm. In this study, the effectiveness of RBFNN-MAX2SAT can be estimated by evaluating the proposed models with testing data sets. The results obtained are analysed using the ratio of satisfied clause (RSC), the root means square error (RMSE), and CPU time. The simulated results suggest that the proposed algorithm is effective in doing MAX2SAT logic programming by analysing the performance by obtaining lower Root Mean Square Error, high ratio of satisfied clauses and lesser CPU time.