Abstract
Central pattern generator (CPG) networks produce characteristic repetitive firing patterns that drive motor motions such as that of the pylorus of the lobster. Maintaining the physiological activity of these networks under various conditions is important. Synaptic plasticity, the tuning of the strengths of synapses, may take place based on the activity of the network. Calculations using a simple model featuring the three main neuron types of the lobster pyloric network CPG have shown that simple activity dependent rules can ensure the physiological activity. The tuning of connection strengths based on the activity of the members in the network is a generic idea that may be tested in other systems as well. In this work we studied tri-membered networks of pulse-coupled Belousov-Zhabotinsky oscillators both in numerical simulations and experiments. We first established the characteristic behaviors of the networks with different topologies. We found that the pyloric network topology is unique because it has a preference for triphasic oscillations in a specific peak order when the coupling strengths are equal and the cycle lengths of the oscillators match. We have found that other behaviors may appear when cycle lengths differ more than 4\\%. After establishing the main characteristics of two tuning mechanisms, we found that a method using a common target activity may be sufficient to secure the preferred triphasic behavior even if the natural activities differ 14\\%. Our findings support that activity dependent modification of interaction strength may ensure the stability of network dynamics in CPGs as well as other systems.