A method by one or more network devices for determining parameter values for a base station of a cellular network operating in a wireless band that is shared with one or more wireless access points. The parameter values are determined to optimize network performance in a manner that allows for fair coexistence between the base station and the one or more wireless access points. The method includes determining, using a machine learning system, parameter values for the base station based on proximity information and activity information for the one or more wireless access points, causing the base station to be configured with the parameter values, determining a measure of how configuring the base station with the parameter values affected network performance based on comparing the level of network performance before and after the base station was configured with the parameter values, and training the machine learning system using the measure.