Deep Learning in Time Series Analysis

8 Deep Time Growing Neural Network

Deep learning was first introduced in the architecture of the multi-layer neural network with a high number of layers and neurons, in a way to improve the learning capacity as was defined in the previous chapters. Time growing neural network was also introduced as an alternative to time-delayed neural network in which dynamic contents of an input time series is incorporated into a set of the temporal window with growing length, each characterized by its frequency contents, in conjunction with a multi-layer perceptron neural network for a nonlinear mapping. Invoking the growing time windows in any form of the growing scheme, can boost the performance of the classification method, compared to the time-delayed ...