Publication: S-Type Random K Satisfiability Logic In Discrete Hopfield Neural Network
Loading...
Date
2025-05
Authors
Osman, Suad Abdeen Helaly
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In the current satisfiability logic representation in artificial neural networks, reliability is crucial for the interpretability of black box models. However, the issue in existing logic is the neglect the probability distribution of the dataset that caused uncertainty. In this thesis, a new satisfiability logical rule namely s-type random 2 satisfiability is proposed with probability distribution to determine the number of order logic clauses and negative literals. The proposed logic includes a probability logic phase that distributes negative literals and ensures the generation of an optimal structure, which achieved the lowest clauses and negative error of 0.41. Subsequently, the hybrid binary sine cosine algorithm with a multi-objective function is proposed in the retrieval phase to obtain a high-quality final neuron state. The proposed algorithm achieved the lowest similarity index of 0.02 when compared to existing algorithms. The new supervised preprocessing phase was proposed to assist in preparing the datasets with different existing methods and demonstrate superior performance. Finally, a new logic mining model was proposed, named s-type random 2 satisfiability based reverse analysis with the new dynamic unit discrete hopfield neural network method and effectively covers the entire solution space to extract all optimal induced logic as compared to existing models. The proposed logic mining model was applied to analyze the non-performing loans dataset and exhibited outstanding performance which achieves an accuracy of 100%.
Description
Keywords
S-Type Random Satisfiability Logic Discrete Hopfield