Kymatio: Scattering Transforms in Python
Mathieu Andreux
(1)
,
Tomás Angles
(2)
,
Georgios Exarchakis
(2)
,
Roberto Leonarduzzi
(2)
,
Gaspar Rochette
(2)
,
Louis Thiry
(2)
,
John Zarka
(2)
,
Stéphane Mallat
(2, 3, 4)
,
Joakim Andén
(5)
,
Eugene Belilovsky
(6)
,
Joan Bruna
(7)
,
Vincent Lostanlen
(7)
,
Matthew J. Hirn
(8)
,
Edouard Oyallon
(9)
,
Sixin Zhang
(10)
,
Carmine Cella
(11)
,
Michael Eickenberg
(3)
1
Owkin France
2 DI-ENS - Département d'informatique - ENS Paris
3 Flatiron Institute
4 Collège de France - Chaire Sciences des données
5 KTH - Department of Mathematics [Sweden]
6 MILA - Montreal Institute for Learning Algorithms [Montréal]
7 NYU - New York University [New York]
8 Michigan State University [Traverse City]
9 MLIA - Machine Learning and Information Access
10 Peking University [Beijing]
11 CNMAT - Center for New Music and Audio Technologies
2 DI-ENS - Département d'informatique - ENS Paris
3 Flatiron Institute
4 Collège de France - Chaire Sciences des données
5 KTH - Department of Mathematics [Sweden]
6 MILA - Montreal Institute for Learning Algorithms [Montréal]
7 NYU - New York University [New York]
8 Michigan State University [Traverse City]
9 MLIA - Machine Learning and Information Access
10 Peking University [Beijing]
11 CNMAT - Center for New Music and Audio Technologies
Vincent Lostanlen
- Fonction : Auteur
- PersonId : 749246
- IdHAL : lostanlen
- ORCID : 0000-0003-0580-1651
- IdRef : 203022769
Edouard Oyallon
- Fonction : Auteur
- PersonId : 179157
- IdHAL : edouard-oyallon
- ORCID : 0000-0002-4826-7527
- IdRef : 228745500
Résumé
The wavelet scattering transform is an invariant signal representation suitable for many signal processing and machine learning applications. We present the Kymatio software package, an easy-to-use, high-performance Python implementation of the scattering transform in 1D, 2D, and 3D that is compatible with modern deep learning frameworks. All transforms may be executed on a GPU (in addition to CPU), offering a considerable speed up over CPU implementations. The package also has a small memory footprint, resulting inefficient memory usage. The source code, documentation, and examples are available undera BSD license at https://www.kymat.io/