newmanrs / streamlit-percolation-model Goto Github PK

Logarithmic cluster size coloring is a poorly thought out idea/feature that only makes the heatmap worse - using fraction of number of possible sites is the correct rendering for the percolation demo. Code will be better with this feature removed.

New versions new problems. Teach me not to pin them.

Clustering algorithm currently only works for when Nx=Ny. This should be fixed, or I should simplify the demo to only allow Nx = Ny and remove code. There isn't really any educational value in the demo supporting rectangles anyhow.

Probably move chart to top and follow up with text, or attempt to make it a two-page app of the simulation and some theory/links of percolation.

Auto play/pause buttons instead of user clicking reset.

Setup free tier deployment with a github action

While N doesn't change, maybe add each hit of "reset" to a scatterplot for max cluster size fraction. Lets the user get a feel for whether the system of does or does not percolate? Might be too messy or noisy, but I think worth prototyping.

Naive brute force clustering algorithm to find largest cluster is currently approximately O(N^3), lagging the app if the simulation is much more than 30x30. Implementing the disjoint set algorithm for clustering should resolve this slow spot.

Or maybe reimplement most of the simulator in C++ and expose python bindings. Maybe overkill for a demo...