sets or evaluation on sets among the evaluated neural networks may cause potentially worthwhile architectures to be rejected prematurely, obviating the advantages realizable by a directed trial and error process. It is an object of this invention to provide a method for improving neural network architectures via an evolutionary algorithm that reduces the adverse effects of the noise that is introduced by the network initialization process. It is a further object of this invention to reduce the noise that is introduced by the network initialization process. It is a further object of this invention to provide an optimized network initialization process. It is a further object of this invention to reduce the noise that is introduced by the use of randomly selected training or evaluation input sets. ;These objects and others are achieved by including parameters that affect the initialization of a neural network architecture within the encoding that is used by an evolutionary algorithm to optimize the neural network architecture. The example initialization parameters include an encoding that determines the initial nodal weights used in each architecture at the commencement of the training cycle. By including the initialization parameters within the encoding used by the evolutionary algorithm, the initialization parameters that have a positive effect on the performance of the resultant evolved network architecture are propagated and potentially improved from
展开▼
