Composite neural network models and their application

A composite neural network is especially suitable for constructing large-scale and heterogeneous neural networks. The large-scale and heterogeneous neural networks are made up of many small-scale networks which are trained individually. The composite neural networks treat these trained networks as components (units) and reuse them as resources. The architecture of each module is characterized by an object-oriented network architecture that facilitates functional network modules and connectionist composition. The object-oriented model contains two primary design concepts, aggregation and generalization, for the description of database objects, aggregate class and the set of instances. We apply these abstraction mechanisms into neural network models as scheme for building large-scale and heterogeneous neural networks. We present a new building tool for constructing the learning space that consists of many separate modular networks, each of which learns to handle a subset of the complete set of training examples.