Multi-tenant BS (MBS) is a new architecture to improve network capacity in which a single BS is tenanted to multiple operators. MBS allows multiple tenants to transmit from a single BS by using their own spectrum resources, resulting in a closer proximity to UEs. In this paper, we propose a UEs to MBSs association scheme, executed in the centralised controller of wireless networks. Each UE can be associated with more than one MBS. We first formulate an optimisation problem that maximises the spectral efficiency of a MBS network with UEs to MBSs associations as its optimisation variables. To solve it, we develop an optimisation solver based on a pursuit learning technique where we model each optimisation variable as a learning automaton. The automata system is executed at the centralised controller where each automaton computes the UE to MBS association in parallel. This results in an extremely low computational complexity and high scalability as compared to other existing schemes. Furthermore, simulation results show that the MBS network has significantly higher spectral efficiency when compared to a single tenant BS network.