An Effective Intelligent Learning Mechanisms based on Wavelet for Pattern Recognition and Data Analytics

Abstract

In recent years, so many intelligent learning mechanisms are developed and proposed for pattern recognition and data analytics. They are improved versions of traditional machine learning approach in the sense of training and prediction. Such class of new intelligent leaning mechanisms namely Deep Learning (DL) and Deep Belief Networks (DBN) are based on Convolutional Neural Networks (CNN). In which multiple hidden layers effectively learns hidden features of the available raw data in hierarchy of features so that they can recombined and aggregate for better learning. This paper presents a novel architecture of convolution neural network that uses wavelet as activation function so as to attain better learning characteristics in multi mode feature extraction approach. In this work, CIFAR, SNAE2, and STL-10 dataset are used for evaluating the performances of Wavelet based Deep Convolutional Neural Network (DCNN).The effectiveness of theoretical implementation is verified through simulation analysis.