Hybrid Neuro-Symbolic Learning Models Integrating Symbolic Reasoning And Data-Driven Approaches

Abstract

Revolutionizing Embedded Systems Debugging: Applying Retrieval-Augmented Generation To Heterogeneous Log Analysis uncertainty and evolving behavior. Neural networks offer strong prediction but limited interpretability, while symbolic systems ensure logic but lack scalability and adaptability. This research proposes an Explainable AI (XAI) using a Hybrid Neuro-Symbolic Learning Model for dynamic environments by integrating data-driven learning with rule-based reasoning. The Dynamic Transient Search Optimized Deep Neural Network (DTSO-DNN) neural component learns deep latent representations from heterogeneous inputs such as sensor and contextual data. DTSO optimizes DNN training by searching for optimal weights and parameters, improving convergence, avoiding local minima, and increasing stability in dynamic data environments. DNN learns complex nonlinear patterns from data and performs feature extraction and prediction from heterogeneous inputs. Data is taken from the Neuro-Symbolic Decision Dataset with sensor features and labels, normalized using Z-score for standardization, and features are extracted using Discrete Wavelet Transform (DWT) to capture temporal patterns and support symbolic reasoning for improved fault diagnosis. Shapley Additive Explanations (SHAP) is used as a post-hoc explainability technique to quantify the contribution of each input feature to individual model predictions in order to improve interpretability. Total Harmonic Distortion (38.4%), Line Outage Recovery Time (2.0 min), Distributed Energy Resource Failure Recovery Time (1.8 min), Load Shedding (39.3 kWh), Renewable Energy Utilization (97.9%), Renewable Curtailment (4.5%), and Computation Time (0.30s per 5-minute interval) are among the experimental results that show the suggested model achieves strong performance and were implemented in MATLAB. When compared to independent neural or symbolic methods, these findings demonstrate increased accuracy, robustness, and explainability.