High Learning Hierarchical Neural Networks

Designing classifiers on high-dimensional learning datasets is an important but difficult task in many artificial intelligence applications. Classifier design often involves learning algorithms of hierarchical neural networks. Deep learning based on the backpropagation method is commonly used to learn hierarchical networks for classification tasks. In this approach, the multi-layer structure of the neural network is defined a priori. According to the proposed high learning, the structure of multi-layer classifiers results from learning data sets based on the principles of separable aggregation.