Learning filter widths of spectral decompositions using wavelets
Published in Advances in Neural Information Processes (NeurIPS), 2018
A novel approach for end-to-end time series classification using neural networks is presented in this work. The wavelet transform is used to learn the optimal spectral decomposition of the time series for any classification task.
Recommended citation: Khan, H., & Yener B. (2018). "Learning filter widths of spectral decompositions using wavelets." Advances in Neural Information Processes (NeurIPS). http://haidark.github.io/files/wavelet_deconv_2018.pdf