A novel pooling layer based on gaussian function with wavelet transform

Abstract

<p>Convolution represent basic layer in the convolutional neural network, but it<br />can result in big size of the data, which may increase the complexity of the<br />network. Different pooling methods are used to perform down sample these<br />data. In this paper, we have proposed a novel pooling method by using<br />Gaussian function to determine the wavelet filter coefficients . At first, the<br />basic statistics are determined for each pool size of the signal, then Gaussian<br />probability distribution function is determined . According to the procedure<br />of extracting the features , three methods are proposed , the first method is<br />used the normalized values of basic statistics as wavelet filter to be<br />multiplied by original signal, the second method used the determined<br />statistics as features of the original signal ,then multiplied it with constant<br />wavelet filter based on Gaussian ,while the third method is similar to first<br />method, except it depend on entire signal instead of each pool size. The<br />proposed methods are combined with other standard methods such as max<br />and pooling. The experiments are performed on different datasets , and the<br />results show that the proposed methods perform or outperform other<br />methods and can increase performance of the (CNN).</p>