Scattering Representations for Recognition

39 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Scattering Networks for Hybrid Representation Learning (2018, methodology): “The method used [9], [10] consists in minimizing the ‘2 reconstruction error between an input image x and a candidate x̂”
Deep Scattering Spectrum (2013, background): “!)ky tk2 = sup ! Rb z(!)jaj2 ?jhj2(t) : Since e(t) is a normalized white noise, e? 1 is a Gaussian process and (t) is a stationary Rayleigh process. With a Gaussian chaos expansion, one can verify [6] that sup !jRb (!)j1 ˇ=4. Applying Lemma 1 to z= and h= ˚ gives E(j(a) ?˚(t)j2) (1 ˇ=4)jaj2 ?j˚j2(t”
Machine learning for data-driven discovery in solid Earth geoscience (2019, background): “The scattering transform is effective, for example, in capturing key properties in multifractal analysis (78) and stratified continuum percolation relevant to representations of sedimentary processes and transport in porous media, respectively.”
A Multiscale Neural Network Based on Hierarchical Matrices (2018, background): “Research using deep neural networks with multiscale architectures has primarily focused on image [4, 11, 13, 42, 52, 58] and video [47] processing.”