Abstract
SHallow REcurrent Decoders (SHRED) provide a deep learning strategy formodeling high-dimensional dynamical systems and/or spatiotemporal data fromdynamical system snapshot observations. PySHRED is a Python package thatimplements SHRED and several of its major extensions, including for robustsensing, reduced order modeling and physics discovery. In this paper, weintroduce the version 1.0 release of PySHRED, which includes data preprocessorsand a number of cutting-edge SHRED methods specifically designed to handlereal-world data that may be noisy, multi-scale, parameterized, prohibitivelyhigh-dimensional, and strongly nonlinear. The package is easy to install,thoroughly-documented, supplemented with extensive code examples, andmodularly-structured to support future additions. The entire codebase isreleased under the MIT license and is available athttps://github.com/pyshred-dev/pyshred.