Quantum-inspired Probabilistic Database Corruption Detection
Roberto Salazar · Parallel Processing Letters · 2023
A crucial problem of machine learning is the management of unclean probabilistic databases. We approach this problem by applying recent results and methods from quantum information to detect a specific class of database corruption. We present a quantifier of the global corruption of the probabilistic database and show its relationship with detection protocols based on generalized Bell inequalities. Furthermore, we show a relation between the noise generating the corruption and information encoded in the database schema. Finally, we discuss how our work indicates a way to export quantum information results to study noise in probabilistic databases.