Neural Net (Pruned 90%)
binary · 36 views
binary
What It Is
Pruned neural network weights (90% zero) --- extreme sparsity typical of compressed models. Zeros encode as byte 0; non-zero weights scaled to [1, 255] preserving sign via abs
Interpretation
Standard analysis sees: heavy-tailed; right-skewed; few distinct values; aperiodic / broadband; low-complexity (predictable, not noise-like); homoskedastic; multifractal. The atlas finds no named structure, but the source is distinctively extreme on SL(2,ℝ) (Thurston):lyapunov_exponent (+2.8z) — beyond what the standard bank predicts for it.
What standard analysis sees
tail heaviness0.95
asymmetry0.98
occupancy0.07
short-range corr0.21
long-range memory0.46
spectral colour0.78
periodicity0.14
complexity0.15
time-irreversibility0.71
volatility clustering0.13
multifractality0.96
dimensionality0.17
nonstationarity0.57
What the atlas adds
Atlas-extreme metrics the standard bank can’t predict for this source
SL(2,ℝ) (Thurston):lyapunov_exponent | +2.8z | bank-miss 1.0σ |
Hodge–Laplacian:source_fraction | -2.3z | bank-miss 1.2σ |
Möbius-S³:spinorial_asymmetry | +2.0z | bank-miss 2.2σ |
Composition
dtypeuint8
range[0, 255]
unique values192 / 16384
mean ± std5.7 ± 22.1
Render Gallery
.png)
Native > phase_plotraw
.png)
Native > timeseriesraw
.png)
Native > triple_latticeraw
.png)
Boltzmann > lag_map
.png)
Boltzmann > spin_grid
.png)
Cayley > default
.png)
Chladni > plate_bottom
.png)
Chladni > plate_top
_(centered)/signed_log_z/Neural_Net_(Pruned_90%).png)
Heisenberg (Nil) (centered) > signed_log_z
_(centered)/xy_path/Neural_Net_(Pruned_90%).png)
Heisenberg (Nil) (centered) > xy_path
.png)
Hodge–Laplacian > default
/barcode/Neural_Net_(Pruned_90%).png)
Inflation (Substitution) > barcode
/d_curve/Neural_Net_(Pruned_90%).png)
Inflation (Substitution) > d_curve
.png)
Information Theory > default
.png)
Julia Set > escape_histogram
.png)
Klein Bottle > default
.png)
Laplacian > default
.png)
Logarithmic Spiral > default
.png)
Möbius-S³ > mobius_cylinder
.png)
Möbius-S³ > s2_base_map
.png)
Ordinal Partition > default
/phi_spectrum/Neural_Net_(Pruned_90%).png)
Penrose (Quasicrystal) > phi_spectrum
.png)
Persistent Homology > barcode
.png)
Persistent Homology > h1_diagram
.png)
Predictability > default
.png)
Recurrence Quantification > default
.png)
Spectral Analysis > default
.png)
Spectral Graph > default
.png)
Spirograph > mode_spectrum
.png)
Spirograph > phasor_trajectory
.png)
Symplectic > flux_grid
.png)
Symplectic > phase_space
/default/Neural_Net_(Pruned_90%).png)
Ulam Spiral (Sacks) > default
/default/Neural_Net_(Pruned_90%).png)
Ulam Spiral (Square) > default
.png)
Visibility Graph > default
.png)
Wavelet Cascade > default
Atlas Position
| Nearest neighbor | Distance | |
|---|---|---|
| Poker Hands | 4.29 | cross-domain |
| Sensor Event Stream | 4.54 | cross-domain |
| Rainfall (ORD Hourly) | 4.59 | cross-domain |
Which Geometries Light Up
2-adic › 2-adic:mean_distance | rank 294/298 | 0.1145 |
Catch24 › Catch24:SB_TransitionMatrix_3ac_sumdiagcov | rank 5/298 | 0.1991 |
E8 Lattice › E8 Lattice:std_profile | rank 2/298 | 8.4853 |
Gottwald-Melbourne › Gottwald-Melbourne:angular_spectral_structure | rank 297/298 | 0.4353 |
Julia Set › Julia Set:escape_entropy | rank 3/298 | 4.4550 |
Moiré › Moiré:moire_invariance_breadth | rank 295/298 | 0.0032 |
Ordinal Partition › Ordinal Partition:markov_mixing | rank 5/298 | 0.9946 |
Sol (Thurston) › Sol (Thurston):z_variance | rank 294/298 | 0.0180 |
← / → within domain · ⇧← / ⇧→ alphabetical · ⇧← / ⇧→ inside an open render = same view across sources