Automatically deploying distributed Access Control Lists (ACLs) in a software-defined network can ensure their internal services and hosts connectivity, security and reliability. ACLs are often deployed in a switch using Ternary Content-Addressable Memory (TCAM). Since TCAM memory is often too limited to store a large ACL, one has to split the lists and distribute the parts on several switches in such a way that every packet travelling from a source to a destination undergoes the required match-action rules. In this paper, we develop and compare three algorithms based on graph theory and Reinforcement Learning (RL) techniques to automatically distribute ACLs across networks switches, while minimizing their TCAM memory occupancy. We compare the three algorithms on several network topologies to evaluate their efficiency in terms of memory occupancy.