NEURAL SYMBOLIC BIG DATA REASONING MODEL FOR EXPLAINABLE AI

Although Artificial Intelligence gained widespread momentum in critical fields like health economics and autonomous technologies it has recently addressed a notable shortcoming in current advanced models: lack of interpretation. Although deep learning models have reached superhuman levels in pattern recognition and prediction tasks within Big Data these models act similarly to "black boxes." They associate inputs with outputs through high-dimensional vector spaces which humans cannot interpret. However symbolic Artificial Intelligence models provide advanced interpretation capabilities through their workings based on symbolic logic and reasoning. These models have not yet scaled well to accommodate uncontrolled Big Data. The Proposed work aims at developing a Neural-Symbolic Big Data Reasoning Model which is a hybrid architecture approach to bridge the gap between the learning ability of neural networks and the reasoning ability offered by symbolic logic. In this approach the connectionist learning aspect offered by neural networks would be combined with symbolic representation capabilities to create a system that would develop an Explainable AI (XAI) methodology with the ability to handle large datasets while at the same time offering clear interpretable and meaningful results to explain their decisions. The Proposed work would make use of deep neuronic networks for carrying out the lower-level features’ extraction with regard to amorphous data forms such as image text or logs which would be mapped to create a symbolic representation graph. A reasoning engine would be utilized to make inferential decisions using symbolic logic ensuring that every predictive result has a traceable decision path which would not only improve trust but also make the system more robust in terms of handling sparse data.