The optimization of neural networks is not necessarily a search process, as the computation involved in teasing out attributes is more of a projection process. While inference may involve some level of search, the training process itself is not based on search.