Neural network based approach for developing embedded software

In parallel with the quick expansion of embedded devices come the rapid growth of embedded software. The commonly used software developing methods can´t meet the fast growing demands of variable software when concerning user experience, cross-platforms, upgrade, hardware resources, etc. This paper proposes a new embedded software developing architecture based on the neural network organization and component principles. With the advantages of the network´s dynamic weight adjusting and flexible self-learning structure, the approach can not only enable the software better adapt to the users´ attributes, but also make the software be easily to extend and upgrade.