Efficient Strategies for Finding the Minimum Information Partition in Integrated Information Theory 3.0

The problem of finding the Minimum Information Partition (MIP) in the context of Integrated Information Theory (IIT) 3.0 presents significant computational challenges due to the complexity of calculating integrated information in large, interconnected systems. Addressing the need for efficient solutions, this work explores bounded approximations to the formalism proposed by IIT 3.0, aiming to establish the theory’s specified properties in more complex systems and to scale with system size and architecture. In this study, we present two strategies aimed at efficiently solving the problem of finding the MIP. The first strategy is based on classical searches with a top-down approach supported by memoization. Beyond classical searches, the second technique introduces a metaheuristic inspired by biological evolution and genetics to reduce processing times and computational complexity. Several test cases for systems of different sizes were conducted, and their results were compared with those presented by the PyPhi application. The proposed optimizations were based on probability properties, approximations using a top-down approach supported by memoization, and metaheuristic techniques aimed at solving combinatorial optimization problems.