Solving Hard XAI Queries Based on a Compiled Dual-Rail Encoding

To address the computational difficulties associated with Boolean classifiers, researchers proposed a solution based on compiled two-rail encoding. This approach utilizes the two-rail encoded representation of classifiers, enabling efficient computation of anti-fact scenario categories that were previously difficult to handle. These categories include shorter anti-fact scenarios and explanations that incorporate the preferences of the subjects being analyzed. The study demonstrated that this method successfully restored the advantages of using compiled representations in dealing with such problems, effectively solving the computational challenges associated with some anti-fact scenario categories in ordered binary decision diagrams.