Convergence vs. Delay Parameter B
This experiment demonstrates how the maximum delay parameter $B$ affects convergence of asynchronous sheaf diffusion. Each agent's update and broadcast periods are sampled from a Gaussian mixture with mean proportional to $B$, so larger $B$ means longer delays between updates.
We run the same diffusion on three graph topologies — a 4-regular graph, an Erdős–Rényi random graph, and a star graph — and sweep $B \in \{1, 10, 50, 100, 200\}$.
Data source: Pre-generated by docs/scripts/acc26_experiments.jl. To regenerate the data and re-render the figure, run:
julia --project=docs docs/scripts/acc26_experiments.jl
julia --project=docs docs/scripts/generate_figures.jlEnergy loss and relative error across graph topologies
Top row: energy $f(\mathbf{x}(t))$ for each topology. Bottom row: relative error $\|\mathbf{x}(t)-\mathbf{x}^*\|/\|\mathbf{x}^*\|$.
