25 July 2020 in Neural Computing and Applications
We address the following problem: given a set of complex images or a large database, the numerical and computational complexity and quality of approximation for neural network may drastically differ from one activation function to another. A general novel methodology, scaled polynomial constant unit activation function “SPOCU,” is introduced and shown to work satisfactorily on a variety of problems. Moreover, we show that SPOCU can overcome already introduced activation functions with good properties, e.g., SELU and ReLU, on generic problems. In order to explain the good properties of SPOCU, we provide several theoretical and practical motivations, including tissue growth model and memristive cellular nonlinear networks. We also provide estimation strategy for SPOCU parameters and its relation to generation of random type of Sierpinski carpet, related to the [pppq] model. One of the attractive properties of SPOCU is its genuine normalization of the output of layers. We illustrate SPOCU methodology on cancer discrimination, including mammary and prostate cancer and data from Wisconsin Diagnostic Breast Cancer dataset. Moreover, we compared SPOCU with SELU and ReLU on large dataset MNIST, which justifies usefulness of SPOCU by its very good performance.
1 September 2008 in Statistics & Probability Letters
19 January 2018 in Neural Computing and Applications
In recent years, air pollution has been a major concern for its implications on human health. Specifically, ozone (\(\mathrm{O}_{3}\)) pollution is causing common respiratory diseases. In this paper, we illustrate the process of modeling and prediction hourly \(\mathrm{O}_ {3}\) pollution measurements using wavelet transforms. We split the time series of \(\mathrm{O}_{3}\) in daily intervals and estimate scale and wavelet coefficients for each interval by the discrete wavelet transform (DWT) with Haar filter. Subsequently we apply cumulated autoregressive integrated moving average (ARIMA) to estimate the coefficients and forecast their evolution in future intervals. Then the inverse discrete wavelet transform is implemented for the reconstruction of the time series and the forecast in the near future. In order to assess the performance of the proposed methodology, we compare the predictions obtained by the DWT–ARIMA with those obtained by the ARIMA model. Several theoretical results are shown through a simulation study.
1 March 2016 in Ecological Complexity