An MDL-Based Wavelet Scattering Features Selection for Signal Classification

Wavelet scattering is a redundant time-frequency transform that was shown to be powerful tool in signal classification. It shares the convolutional architecture with neural networks, but it offers some advantages, including faster training and small sets. However, introduces redundancy along frequency axis, especially for filters have high degree of overlap. This naturally leads need dimensionality reduction further increase its efficiency as machine learning tool. In this paper, Minimum Description Length used define an automatic procedure optimizing selection features, even domain. The proposed study limited class uniform sampling models. Experimental results show method able automatically select optimal step guarantees highest classification accuracy fixed parameters, when applied audio/sound signals.