c) machine learning model - EWSNet
The final EWSmethods module is an interface to the Python based
EWSNet, a deep learning modelling framework for predicting critical
transitions and tipping points (Deb et al. 2022). EWSNet consists of
coupled long short-term memory and fully convolutional network
sub-module routines, which together extract complex nonlinear patterns
from inputted time series to provide forecasts on the likelihood of
oncoming tipping points. Details on the precise formulation and model
structure can be found at Deb et al. (2022) and
https://ewsnet.github.io, whereas here we will focus on the
application of EWSNet for ecologists and the setup of the R environment
to cooperate with EWSNet’s Python backend.