A high order Boltzmann machine configuration for transfer sequence searches in decomposed state machines

Summary form only given, as follows. A neural network architecture in the form of a high-order Boltzmann machine is proposed for the transfer sequence search of a decomposed sequential machine. The transfer sequence for each component machine can be found using a second-order Boltzmann machine, but the dependencies between component machines require high-order connections. The Boltzmann machine configuration as a constraint satisfaction network for a sequential machine structure is defined. The global and the local energy terms for the overall architecture are defined based on the component machine state tables and the state transition dependencies between machines. The application of the proposed architecture to a sample problem is demonstrated by using a fourth-order Boltzmann machine for the transfer sequence search in a serially decomposed sequential machine.<>