RANDU

algorithmic-bytes · 37 views
algorithmic-byteshigh-entropy

What It Is

IBM's infamous RANDU (1968) --- the worst widely-deployed RNG in history. Points fall on just 15 parallel planes in 3D. Multiplier 65539 = 2¹⁶+3

Interpretation

Standard analysis sees: rich, high-entropy values; aperiodic / broadband; high-complexity (noise-like); high-dimensional / space-filling. The atlas detects no named structure beyond this. It sits beside Wichmann-Hill in the atlas (standard-bank rank 21) — a neighbor conventional features miss.

What standard analysis sees
tail heaviness0.22
asymmetry0.39
occupancy0.93
short-range corr0.33
long-range memory0.35
spectral colour0.71
periodicity0.04
complexity0.92
time-irreversibility0.26
volatility clustering0.17
multifractality0.20
dimensionality0.96
nonstationarity0.24
What the atlas adds

Nothing beyond the standard reading — this source’s structure is already captured by standard features; the atlas adds no named residual.

Composition

dtypefloat64
range[2.283e-05, 0.9998]
unique values16384 / 16384
mean ± std0.499 ± 0.287

Render Gallery

Atlas Position

Nearest neighborDistance
XorShift321.67
ChaCha201.68
Wichmann-Hill1.95

Open in Atlas →

Which Geometries Light Up

Catch24Catch24:SB_MotifThree_quantile_hhrank 5/3072.1971
Modular ResidueModular Residue:equidistribution_deviationrank 305/3070.0022
Mostow RigidityMostow Rigidity:volume_sum_jitter_stabilityrank 3/3070.9676
PredictabilityPredictability:cond_entropy_k1rank 5/3072.9976
in stochastic-process
alphabetical
← / → within domain · ⇧← / ⇧→ alphabetical · ⇧← / ⇧→ inside an open render = same view across sources