Survey on Optimization of Number of Hidden Layers in Neural Networks

Abstract: Artificial neural networks attempt to mimic the basic operations of the brain. Information is pass between the neurons, and based upon the structure and synapse weights, a network behavior or output mapping is provided. Challenges in artificial neural network are finding number of hidden layer and hidden neurons in respected layer, design of a network with more hidden layers, designing optimal network, finding minimal network structure in less time, determination of processing and storage resources in implementation of large and effective network. Among these challenges, finding number of hidden layers and hidden neurons in the respected layer is the core objective of this system. Multi layer neural network permit more complex, non linear relationship of input data to output results. Hidden layer is the intermediate unit between input and output unit used to calculate weighted sums of the inputs. The output result of the network is dependent on hidden layer result. In a network with a hidden layer and an output layer, the hidden layer is computed first and then the result of the hidden layer are used to compute the output layer.

Keywords: Artificial Neural Networks; hidden neuron; hidden layer.