A software development process model for artificial neural networks in critical applications

Using artificial neural networks (ANNs) in critical applications can be challenging due to the often experimental nature of ANN construction and the "black box" label that is frequently attached to ANNs. Well-accepted process models exist for algorithmic software development which facilitate software validation and acceptance. The software development process model presented herein is targeted specifically toward ANNs in critical applications. This model is not unwieldy, and could easily be used on projects without critical aspects. This should be of particular interest to organizations that use ANNs and need to maintain or achieve a capability maturity model or ISO software development rating. Further, while this model is aimed directly at neural network development, with minor modifications, the model could be applied to any technique wherein knowledge is extracted from existing data, such as other numeric approaches or knowledge-based systems.