Addition of mean_adjacent_gap_ratio.

Author: msuzen

NEWS

@@ -1,3 +1,7 @@
+v0.2.32
+
+* Addition of `mean_adjacent_gap_ratio` and test.
+
 v0.2.22
 
 * NNSD with spectral unfolding. Implemented and tested.

README.md

@@ -47,8 +47,9 @@ The core package is providing strong randomness improving the simulation quality
 * Generation of Mixed Gaussian ensembles (Orthogonal) via `Mixed Matrix Ensemble Sampling (MMES) algoritm`
 * Extract offdiagonal elements.
 * Spectra generation given ensemble.
-* Spectral unfolding.
+* Robust Spectral unfolding.
 * Nearest-Neigbour Spacing Densities (NNSD).
+* Adjacent gap ratio.
 * Analytic distributions: Wigner semi-circle law, nearest-neigbour spacing.
 
 ### Statistics

leymosun/spectral.py

@@ -185,3 +185,25 @@ def nnsd(eigenvalues: np.array, locations: np.array = np.arange(0.0, 5.0, 0.1)):
         density, _locations = pdf(ueigenvalues_nn, locations=locations)
         nnsd_densities.append(density)
     return np.array(nnsd_densities), np.array(_locations)
+
+
+def mean_adjacent_gap_ratio(eigenvalues):
+    """Compute ratio of adjacent spacing gap ratio
+       Due to Oganesyan-Huse Phys. Rev. B 75, 155111 (2007)
+
+    Args:
+        eigenvalues: 2D, eigenvalues over repeaded samples.
+        locations: Centers to compute PDF of NNSD.
+
+    Returns:
+        Mean given ensemble of eigenvalues
+
+    """
+    adjacent_gap_ratios = []
+    for eigens in  eigenvalues:
+        eigens_sorted = np.sort(eigens)
+        spacings = np.diff(eigens_sorted)
+        adjacent_gap_ratio = np.minimum(spacings[:-1], spacings[1:]) /(np.maximum(spacings[:-1], spacings[1:])+1e-9)
+        adjacent_gap_ratios.append(np.mean(adjacent_gap_ratio))
+    bar_adjacent_gap_ratio = np.mean(adjacent_gap_ratios)
+    return bar_adjacent_gap_ratio

leymosun/version.py

@@ -1 +1 @@
-__version__="0.2.22"
+__version__="0.2.32"

setup.py

@@ -17,7 +17,8 @@
     description="High-Entropy Randomness Research Toolkit. High-Entropy Random Number Generator (HE-RNGs). \
                 Generation of random matrices in canonical ensembles. Generating of mixed matrix ensembles via \
                 Mixed Matrix Ensemble Sampling (MMES) algorithm. \
-                Matrix utilities, spectral analysis tools, analytic expressions from theory and \
+                Matrix utilities, spectral analysis, spectral unfolding, nearest-neighbor spacing, \
+                mean adjacent gap ratio, analytic expressions from theory and \
                 uncertainty quantification tools.",
     long_description=long_description,
     long_description_content_type="text/markdown",

tests/test_spectral.py

@@ -4,7 +4,8 @@
     empirical_spectral_density,
     eigenvalue_on_polynomial,
     unfold_spectra,
-    nnsd
+    nnsd,
+    mean_adjacent_gap_ratio
 )
 from leymosun.matrix import ensemble, mixed_ensemble
 from leymosun.gaussian import goe
@@ -60,3 +61,8 @@ def test_nnsd():
     sim_e2 = np.array([normal(size=1000) for _ in range(10)])
     nnsd_densities, locations = nnsd(sim_e2)
     assert nnsd_densities.shape[1] == 48
+    
+def test_mean_adjacent_gap_ratio():
+    sim_e2 = np.array([normal(size=1000) for _ in range(10)])
+    r = mean_adjacent_gap_ratio(sim_e2)
+    assert r > 0.0