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.