We present a computational model to study the robustness and degradation of dynamics on a network that includes a large number of units and connections between them. Each unit has an internal structure and it is connected to other units through contact points. These contact points correspond to the synapses of the biological neural networks. We monitor the network activity as a function of time, after we initiate an input signal at random in the network. We vary the number of connections (as a function of several properties of each connection), and observe that there exists a critical crossover value regarding the loss of connections below which all network activity decreases at a much faster rate than the expected normal loss. This crossover value is in the range of 70–80% loss. A similar critical value observed in biological neural networks may define the limit between the healthy state and the disease. Correlations between the computational and the biological model are discussed.
