This repo uses Jacobian descent to optimize simple multidimensional functions and plots the obtained optimization trajectories.
uv python install 3.13.3
uv python pin 3.13.3./setup_env.sh mainNote that here, "main" can be replaced with whatever ref (branch, tag or commit hash) of torchjd you want.
Alternatively, you may want to install using the uv.lock file to reproduce an exact environment.
You might also need some tex packages to be able to generate the plots (see https://stackoverflow.com/a/53080504)
sudo apt-get install texlive-latex-extra texlive-fonts-recommended dvipng cm-superPlease refer to the docstring of the scripts.
Optimization of the function f(x₁, x₂) = [x₁², x₂²]ᵀ by various aggregators.
Trajectories in the parameter space:
Trajectories in the value space:
Optimization of the function described in Eq. 14 of https://arxiv.org/pdf/2406.16232v3
Trajectories in the parameter space:
Trajectories in the value space:
