Generally, the resolution of a problem by using soft-computing support requires several attempts for setting up a proper neural network. Such attempts consist of designing and training a neural network and can be a relevant effort for the developer. This paper proposes a toolbox that automates several steps for setting up a neural network, and provides high-level abstractions allowing a developer to choose classical network topologies and configure them as desired, as well as design a neural network from a scratch. A valuable aspect of our solution is given by the modularity of the whole design that builds on object-orientation and design patterns.
An Object-Oriented Neural Network Toolbox Based on Design Patterns
NAPOLI, CHRISTIAN;TRAMONTANA, EMILIANO ALESSIO
2015-01-01
Abstract
Generally, the resolution of a problem by using soft-computing support requires several attempts for setting up a proper neural network. Such attempts consist of designing and training a neural network and can be a relevant effort for the developer. This paper proposes a toolbox that automates several steps for setting up a neural network, and provides high-level abstractions allowing a developer to choose classical network topologies and configure them as desired, as well as design a neural network from a scratch. A valuable aspect of our solution is given by the modularity of the whole design that builds on object-orientation and design patterns.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.