Kymatio: Scattering Transforms in Python

Andreux, Mathieu; Angles, Tomás; Exarchakis, Georgios; Rochette, Gaspar; Thiry, Louis; Zarka, J.; Mallat, Stéphane; Andén, Joakim; Belilovsky, Eugene; Bruna, Joan; Lostanlen, Vincent; Chaudhary, Muawiz; Hirn, Matthew; Oyallon, Edouard; Zhang, Sixin; Cella, Carmine Emanuele; Eickenberg, Michael

doi:10.48550/arxiv.1812.11214

Cited by 13 publications

(23 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…, based on the particular 3D implementation of Ref. [42], given by the publicly available KYMATIO package [43] 4 . In Eq.…”

Section: Wavelet Scattering Transformmentioning

confidence: 99%

“…As opposed to previous cosmological applications of wavelet techniques [39][40][41] that restricted their attention on 2D configurations only, in this work we aim to study the information content of the WST applied on the full 3D density field of the LSS. To do so, we will closely follow the corresponding 3D WST implementation of [42], given by the publicly available KYMATIO package [43] 1 , based on which the WST coefficients, S n , up to order n = 2, are given by:…”

mentioning

confidence: 99%

See 1 more Smart Citation

Towards an Optimal Estimation of Cosmological Parameters with the Wavelet Scattering Transform

Valogiannis,

Dvorkin

2021

Preprint

View full text Add to dashboard Cite

“…, based on the particular 3D implementation of Ref. [42], given by the publicly available KYMATIO package [43] 4 . In Eq.…”

Section: Wavelet Scattering Transformmentioning

confidence: 99%

mentioning

confidence: 99%

Towards an Optimal Estimation of Cosmological Parameters with the Wavelet Scattering Transform

Valogiannis,

Dvorkin

2021

Preprint

View full text Add to dashboard Cite

“…Over the years, a number of software packages have been developed to implement the scattering transform. Recently, the developers of several of those teamed up to create the "Kymatio" package (Andreux et al 2018): https://www.kymat.io. It can take advantage of GPU acceleration, provides a variety of different frontends and backends, and has been widely used in the signal processing literature.…”

Section: Software Implementationmentioning

confidence: 99%

How to quantify fields or textures? A guide to the scattering transform

Cheng¹,

Ménard²

2021

Preprint

View full text Add to dashboard Cite

Extracting information from stochastic fields or textures is a ubiquitous task in science, from exploratory data analysis to classification and parameter estimation. From physics to biology, it tends to be done either through a power spectrum analysis, which is often too limited, or the use of convolutional neural networks (CNNs), which require large training sets and lack interpretability. In this paper, we advocate for the use of the scattering transform (Mallat 2012), a powerful statistic which borrows mathematical ideas from CNNs but does not require any training, and is interpretable. We show that it provides a relatively compact set of summary statistics with visual interpretation and which carries most of the relevant information in a wide range of scientific applications. We present a non-technical introduction to this estimator and we argue that it can benefit data analysis, comparison to models and parameter inference in many fields of science. Interestingly, understanding the core operations of the scattering transform allows one to decipher many key aspects of the inner workings of CNNs.

show abstract

“…We used many publicly available codes, including PYTHON packages: HDF5 [61], IPYTHON [62], KYMATIO [52], MAT-PLOTLIB [63], NUMPY [64], SCIPY [65], SCIKIT-LEARN [66], SCIKIT-IMAGE [67], DYNESTY [68] and JULIA [69] packages: FFTW.jl [70], AbstractFFTs.jl, Colors.jl, DSP.jl, FITSIO.jl, HDF5.jl, IJULIA.jl, MLDatasets.jl, and would like to acknowledge their developers. ACKNOWLEDGMENT A.S. gratefully acknowledges support by a National Science Foundation Graduate Research Fellowship (DGE-1745303).…”

Section: Code and Data Availabilitymentioning

confidence: 99%

“…Note that here we have only one free parameter, α of the apodization function, which replaces the free parameters associated with that spherical projection.10 KYMATIO[52] is a PYTHON implementation of the WST available at https://www.kymat.io11 https://faun.rc.fas.harvard.edu/saydjari/EqWS/12 The full width at half maximum (FWHM) is a measure of the width, equal to 2.355σ, which we report in pixel units.…”

mentioning

confidence: 99%