Procedural generator

How it works

Candidates come from a stochastic generator (smoothed Gaussian-random-field topologies plus strut/lattice motifs), optionally folded to enforce symmetry. Each is scored by the same homogenization algorithm used to label the project dataset, ported to JavaScript and validated against the Python solver.

Auxetic note: Negative Poisson's ratio needs slender, well-proportioned re-entrant lattices that a coarse pixel grid can't represent, so the "Min Poisson" goal searches for the lowest value it can find and reports the real number. It is usually still positive.