SA Levy Flight (1D) + PERM-like Enrichment - Demo - Little Fluctuating Cat _(:з」∠)

Controls

σ (tail exponent)

1.50

N_max (steps per tour) r_max (truncation)

c_low (prune threshold)

c_high (clone threshold)

tours per second (speed)

20 tours/s

This visualizes one tour at a time as a branching population (PERM-like prune/clone). The statistics panel aggregates over tours.

Note: in the provided Python, weights w stay ~1 (no Rosenbluth bias), so prune/clone rarely triggers. This demo keeps the same behavior, but you can force more pruning/cloning by setting c_low close to 1 and c_high close to 1.

Visualization + Live Stats

history (faded) active chains cloned pruned

Tours completed

Current step n

Active chains

p_n (proxy)

—

⟨log R_n⟩

—

R_geom

—

Scaling reminders (asymptotic): if R_geom(n) ~ n^ν, then ν ≈ d⟨log R⟩/d log n. If p_n ~ μ^n n^{γ-1}, then A_n=log(p_n/p_{n-1}) ≈ log μ + (γ-1)/n.