Solving HornSAT Fuzzy Logic Neuro-symbolic Integration

Abstract

This research proposes a new method for improving logic programming efficiency in the Hopfield network. Basic logic program alone without integration with other computing networks have limits when modelling datasets due to its inability to capture all the data collection characteristics. An improved logical rule is required to retrieve the data sets’ hidden information. Hence, a new model of integrating fuzzy logic with HornSAT in the Hopfield network is proposed to represent information more effectively and in a comprehensive logical rule. The combination of fuzzy logic and HornSAT can deal with the combinatorial optimization problems in Hopfield networks. Fuzzification and defuzzification processes can reduce the computing load to find the correct states of the neurons. The proposed approach can make the selection of neuron state wider between 0 and 1 in solving the HornSAT problem. Aside from that, during the defuzzification process, unsatisfied neuron clauses will be adjusted using the alpha-cut method until the correct neuron state is identified. From the results shown, the proposed method proves to enhance the Hopfield network performance with effective results in terms of the global minimum ratio, hamming distance and processing period. The conclusion is fuzzy logic HornSAT increases the network performance better than the Direct technique.