Abstract. Artificial neural networks are intended to be used in future nanoelectronics since their biological examples seem to be robust to noise. In this paper, we analyze the robustness of Local Cluster Neural Networks and determine upper bounds on the mean square error for noise contaminated weights and inputs.