freealg 0.1.7__tar.gz → 0.1.8__tar.gz

{freealg-0.1.7 → freealg-0.1.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: freealg
-Version: 0.1.7
+Version: 0.1.8
 Summary: Free probability for large matrices
 Keywords: leaderboard bot chat
 Platform: Linux
@@ -30,7 +30,6 @@ Requires-Dist: scipy
 Requires-Dist: texplot
 Requires-Dist: matplotlib
 Requires-Dist: colorcet
-Requires-Dist: networkx
 Requires-Dist: statsmodels
 Provides-Extra: test
 Requires-Dist: tox; extra == "test"

freealg-0.1.8/freealg/__version__.py

@@ -0,0 +1 @@
+__version__ = "0.1.8"

{freealg-0.1.7 → freealg-0.1.8}/freealg/distributions/kesten_mckay.py

@@ -12,7 +12,6 @@
 # =======
 
 import numpy
-import networkx as nx
 from scipy.interpolate import interp1d
 from .._plot_util import plot_density, plot_hilbert, plot_stieltjes, \
     plot_stieltjes_on_disk, plot_samples
@@ -450,7 +449,7 @@ class KestenMcKay(object):
             is used.
 
         method : {``'mc'``, ``'qmc'``}, default= ``'qmc'``
-            Method of drawing samples from uniform distirbution:
+            Method of drawing samples from uniform distribution:
 
             * ``'mc'``: Monte Carlo
             * ``'qmc'``: Quasi Monte Carlo
@@ -534,6 +533,39 @@ class KestenMcKay(object):
 
         return samples
 
+    # ============
+    # Haar unitary
+    # ============
+
+    def _haar_orthogonal(self, n, k):
+        """
+        Haar-distributed O(n) via the Mezzadri QR trick.
+
+        References
+        ----------
+
+        .. [1] Francesco Mezzadri. How to generate random matrices from the
+               classical compact groups. https://arxiv.org/pdf/math-ph/0609050
+
+        Notes
+        -----
+
+        Taking the QR of a normal-Gaussian matrix gives an orthonormal basis,
+        but the columns of that Q are not uniform on the sphere, as they are
+        biased by the signs or phases in the R-factor.
+
+        With Mezzadri method, columns of Q are rescaled by the reciprocals of
+        the diagonals of R phase, resulting in a matrix that is exactly
+        uniformly distributed under Haar measure O(n).
+        """
+
+        rng = numpy.random.default_rng()
+        Z = rng.standard_normal((n, k))
+        Q, R = numpy.linalg.qr(Z, mode='reduced')   # Q is n by k
+        Q *= numpy.sign(numpy.diag(R))
+
+        return Q
+
     # ======
     # matrix
     # ======
@@ -554,6 +586,43 @@ class KestenMcKay(object):
         A : numpy.ndarray
             A matrix of the size :math:`n \\times n`.
 
+        Notes
+        -----
+
+        If the parameter :math:`d` is even, the matrtix is generated from
+
+        .. math::
+
+            \\mathbf{A} = \\sum_{i=1}^{d/2} \\mathbf{O}_i +
+            \\mathbf{O}_o^{\\intercal},
+
+        where :math:`\\mathbf{O}_i` are randomly generated orthogonal matrices
+        with Haar. This method is fast but :math:`d` has to be even.
+
+        If all other :math:`d`, the following is used:
+
+        .. math::
+
+            \\mathbf{A} = \\mathbf{P} \\mathbf{O} \\mathbf{D} \\mathbf{O}^{-1}
+            \\mathbf{P},
+
+        where :math:`\\mathbf{D}` is diagonal matrix with entries
+        :math:`\\pm 1`, :math:`\\mathbf{O}` is orthogonal with Haar measure,
+        and :math:`\\mathbf{P}` is a projection matrix. For more details, see
+        Section 5 and 6 of [1]_.
+
+        The orthogonal matrices are genrated using the method of [2]_.
+
+        References
+        ----------
+
+        .. [1] Iris S. A. Longoria and James A. Mingo, Freely Independent Coin
+               Tosses, Standard Young Tableaux, and the Kesten--McKay Law.
+               https://arxiv.org/abs/2009.11950
+
+        .. [2] Francesco Mezzadri. How to generate random matrices from the
+               classical compact groups. https://arxiv.org/pdf/math-ph/0609050
+
         Examples
         --------
 
@@ -564,11 +633,38 @@ class KestenMcKay(object):
             >>> A = km.matrix(2000)
         """
 
-        n = size
-        G = nx.random_regular_graph(self.d, n)
-        A = nx.to_numpy_array(G, dtype=float)  # shape (n,n)
-
-        mu = self.d / n
-        A_c = A - mu * numpy.ones((n, n))
-
-        return A_c
+        if (self.d >= 2) and (self.d % 2 == 0):
+            # Uses algorithm 1 . Only if d is even. This is much faster than
+            # algorithm 2.
+            n = size
+            rng = numpy.random.default_rng()
+            m = self.d // 2
+            A = numpy.zeros((n, n))
+
+            for _ in range(m):
+                O_ = self._haar_orthogonal(n, n)
+                A += O_ + O_.T
+        else:
+            # Uses algorithm 2. Only when d is odd, but this algorithm works
+            # for any d (even and odd), but it takes much longer to comute
+            # especially if d is larger. As such, as only use algorithm 1 when
+            # d is even and use algorithm 2 for the rest.
+            n = size * self.d
+            rng = numpy.random.default_rng()
+
+            # Deterministic pieces
+            k = size
+            if k == 0:
+                raise ValueError('Choose size larger then d.')
+
+            # Projection rows of O
+            Q = self._haar_orthogonal(n, k)
+            O_k = Q.T
+
+            # diagonal D with equal \pm 1 (trace 0)
+            diag = numpy.ones(n, dtype=float)
+            diag[:n//2] = -1
+            rng.shuffle(diag)
+            A = (n/k) * (O_k * diag) @ O_k.T
+
+        return A

{freealg-0.1.7 → freealg-0.1.8}/freealg/distributions/marchenko_pastur.py

@@ -456,7 +456,7 @@ class MarchenkoPastur(object):
             is used.
 
         method : {``'mc'``, ``'qmc'``}, default= ``'qmc'``
-            Method of drawing samples from uniform distirbution:
+            Method of drawing samples from uniform distribution:
 
             * ``'mc'``: Monte Carlo
             * ``'qmc'``: Quasi Monte Carlo

{freealg-0.1.7 → freealg-0.1.8}/freealg/distributions/meixner.py

@@ -71,7 +71,7 @@ class Meixner(object):
         \\frac{4(1+b) - (x-a)^2}{2 \\pi (b x^2 + a x + 1)}
         \\mathbf{1}_{x \\in [\\lambda_{-}, \\lambda_{+}]} \\mathrm{d}{x}
 
-    where :math:`a, b` are the shape parameters of the distributon. The edges
+    where :math:`a, b` are the shape parameters of the distribution. The edges
     of the support are
 
     .. math::
@@ -98,15 +98,16 @@ class Meixner(object):
     # init
     # ====
 
-    def __init__(self, a, b):
+    def __init__(self, a, b, c):
         """
         Initialization.
         """
 
         self.a = a
         self.b = b
-        self.lam_p = self.a + 2.0 * numpy.sqrt(1.0 + self.b)
-        self.lam_m = self.a - 2.0 * numpy.sqrt(1.0 + self.b)
+        self.c = c
+        self.lam_p = self.a + 2.0 * numpy.sqrt(self.b)
+        self.lam_m = self.a - 2.0 * numpy.sqrt(self.b)
         self.support = (self.lam_m, self.lam_p)
 
     # =======
@@ -172,9 +173,19 @@ class Meixner(object):
         rho = numpy.zeros_like(x)
         mask = numpy.logical_and(x > self.lam_m, x < self.lam_p)
 
-        rho[mask] = \
-            numpy.sqrt(4.0 * (1.0 + self.b) - (x[mask] - self.a)**2) / \
-            (2.0 * numpy.pi * (self.b * x[mask]**2 + self.a * x[mask] + 1))
+        # rho[mask] = \
+        #     numpy.sqrt(4.0 * (1.0 + self.b) - (x[mask] - self.a)**2) / \
+        #     (2.0 * numpy.pi * (self.b * x[mask]**2 + self.a * x[mask] + 1))
+
+        numer = numpy.zeros_like(x)
+        denom = numpy.ones_like(x)
+        numer[mask] = self.c * numpy.sqrt(4.0 * self.b - (x[mask] - self.a)**2)
+        denom[mask] = (1 - self.c)*(x[mask] - self.a)**2
+        denom[mask] += self.a * (2 - self.c)*(x[mask] - self.a)
+        denom[mask] += self.a**2 + self.b * self.c**2
+        denom[mask] *= 2 * numpy.pi
+
+        rho[mask] = numer[mask] / denom[mask]
 
         if plot:
             plot_density(x, rho, label='', latex=latex, save=save)
@@ -239,12 +250,17 @@ class Meixner(object):
             x = numpy.linspace(x_min, x_max, 500)
 
         def _P(x):
-            denom = 1.0 + self.b
-            return ((1.0 + 2.0 * self.b) * x + self.a) / denom
+            # denom = 1.0 + self.b
+            # return ((1.0 + 2.0 * self.b) * x + self.a) / denom
+            P = ((self.c - 2.0) * x - self.a * self.c) / 2.0
+            return P
 
         def _Q(x):
-            denom = 1.0 + self.b
-            return (self.b * x**2 + self.a * x + 1.0) / denom
+            # denom = 1.0 + self.b
+            # return (self.b * x**2 + self.a * x + 1.0) / denom
+            Q = ((1.0 - self.c) * x**2 + self.a * self.c * x +
+                 self.b * self.c**2) / 4.0
+            return Q
 
         P = _P(x)
         Q = _Q(x)
@@ -272,21 +288,32 @@ class Meixner(object):
         """
 
         sign = -1 if alt_branch else 1
-        denom = 1.0 + self.b
-        A = (self.b * z**2 + self.a * z + 1.0) / denom
-        B = ((1.0 + 2.0 * self.b) * z + self.a) / denom
-        D = B**2 - 4 * A
-        sqrtD = numpy.sqrt(D)
+        # denom = 1.0 + self.b
+        # A = (self.b * z**2 + self.a * z + 1.0) / denom
+        # B = ((1.0 + 2.0 * self.b) * z + self.a) / denom
+        A = ((1.0 - self.c) * z**2 + self.a * self.c * z +
+             self.b * self.c**2) / 4.0
+        B = ((self.c - 2.0) * z - self.a * self.c) / 2.0
+
+        # D = B**2 - 4 * A
+        # sqrtD = numpy.sqrt(D)
+
+        # Avoid numpy picking the wrong branch
+        d = 2 * numpy.sqrt(1.0 + self.b)
+        r_min = self.a - d
+        r_max = self.a + d
+        sqrtD = numpy.sqrt(z - r_min) * numpy.sqrt(z - r_max)
+
         m1 = (-B + sqrtD) / (2 * A)
         m2 = (-B - sqrtD) / (2 * A)
 
         # pick correct branch only for non‑masked entries
         upper = z.imag >= 0
         branch = numpy.empty_like(m1)
-        branch[upper] = numpy.where(sign*m1[upper].imag > 0, m1[upper],
-                                    m2[upper])
-        branch[~upper] = numpy.where(sign*m1[~upper].imag < 0, m1[~upper],
-                                     m2[~upper])
+        branch[upper] = numpy.where(
+            sign*m1[upper].imag > 0, m1[upper], m2[upper])
+        branch[~upper] = numpy.where(
+            sign*m1[~upper].imag < 0, m1[~upper], m2[~upper])
         m = branch
 
         return m
@@ -454,7 +481,7 @@ class Meixner(object):
             is used.
 
         method : {``'mc'``, ``'qmc'``}, default= ``'qmc'``
-            Method of drawing samples from uniform distirbution:
+            Method of drawing samples from uniform distribution:
 
             * ``'mc'``: Monte Carlo
             * ``'qmc'``: Quasi Monte Carlo
@@ -571,14 +598,4 @@ class Meixner(object):
             >>> A = mx.matrix(2000)
         """
 
-        n = size
-        m1 = int(self.a * n)
-        m2 = int(self.b * n)
-
-        X = numpy.random.randn(n, m1)
-        Y = numpy.random.randn(n, m2)
-
-        Sx = X @ X.T
-        Sy = Y @ Y.T
-
-        return Sx, Sy
+        raise NotImplementedError

{freealg-0.1.7 → freealg-0.1.8}/freealg/distributions/wachter.py

@@ -456,7 +456,7 @@ class Wachter(object):
             is used.
 
         method : {``'mc'``, ``'qmc'``}, default= ``'qmc'``
-            Method of drawing samples from uniform distirbution:
+            Method of drawing samples from uniform distribution:
 
             * ``'mc'``: Monte Carlo
             * ``'qmc'``: Quasi Monte Carlo

{freealg-0.1.7 → freealg-0.1.8}/freealg/distributions/wigner.py

@@ -12,7 +12,6 @@
 # =======
 
 import numpy
-import networkx as nx
 from scipy.interpolate import interp1d
 from .._plot_util import plot_density, plot_hilbert, plot_stieltjes, \
     plot_stieltjes_on_disk, plot_samples
@@ -77,20 +76,20 @@ class Wigner(object):
     .. code-block:: python
 
         >>> from freealg.distributions import Wigner
-        >>> wg = Wigner()
+        >>> wg = Wigner(1)
     """
 
     # ====
     # init
     # ====
 
-    def __init__(self):
+    def __init__(self, r):
         """
         Initialization.
         """
-
-        self.lam_p = 2.0
-        self.lam_m = -2.0
+        self.r = r
+        self.lam_p = self.r
+        self.lam_m = -self.r
         self.support = (self.lam_m, self.lam_p)
 
     # =======
@@ -136,7 +135,7 @@ class Wigner(object):
         .. code-block::python
 
             >>> from freealg.distributions import Wigner
-            >>> wg = Wigner()
+            >>> wg = Wigner(1)
             >>> rho = wg.density(plot=True)
 
         .. image:: ../_static/images/plots/wg_density.png
@@ -156,8 +155,8 @@ class Wigner(object):
         rho = numpy.zeros_like(x)
         mask = numpy.logical_and(x >= self.lam_m, x <= self.lam_p)
 
-        rho[mask] = (1.0 / (2.0 * numpy.pi)) * \
-            numpy.sqrt(4.0 - x[mask]**2)
+        rho[mask] = (2.0 / (numpy.pi * self.r**2)) * \
+            numpy.sqrt(self.r**2 - x[mask]**2)
 
         if plot:
             plot_density(x, rho, label='', latex=latex, save=save)
@@ -204,7 +203,7 @@ class Wigner(object):
         .. code-block::python
 
             >>> from freealg.distributions import Wigner
-            >>> wg = Wigner()
+            >>> wg = Wigner(1)
             >>> hilb = wg.hilbert(plot=True)
 
         .. image:: ../_static/images/plots/wg_hilbert.png
@@ -225,7 +224,7 @@ class Wigner(object):
             return x
 
         def _Q(x):
-            return 1.0
+            return (self.r**2) / 4.0
 
         P = _P(x)
         Q = _Q(x)
@@ -256,7 +255,7 @@ class Wigner(object):
 
         # Use quadratic form
         sign = -1 if alt_branch else 1
-        A = 1.0
+        A = (self.r**2) / 4.0
         B = z
         D = B**2 - 4 * A
         sqrtD = numpy.sqrt(D)
@@ -346,7 +345,7 @@ class Wigner(object):
         .. code-block:: python
 
             >>> from freealg.distributions import Wigner
-            >>> wg = Wigner()
+            >>> wg = Wigner(1)
             >>> m1, m2 = wg.stieltjes(plot=True)
 
         .. image:: ../_static/images/plots/wg_stieltjes.png
@@ -437,7 +436,7 @@ class Wigner(object):
             is used.
 
         method : {``'mc'``, ``'qmc'``}, default= ``'qmc'``
-            Method of drawing samples from uniform distirbution:
+            Method of drawing samples from uniform distribution:
 
             * ``'mc'``: Monte Carlo
             * ``'qmc'``: Quasi Monte Carlo
@@ -471,7 +470,7 @@ class Wigner(object):
         .. code-block::python
 
             >>> from freealg.distributions import Wigner
-            >>> wg = Wigner()
+            >>> wg = Wigner(1)
             >>> s = wg.sample(2000)
 
         .. image:: ../_static/images/plots/wg_samples.png
@@ -553,15 +552,7 @@ class Wigner(object):
 
         # Parameters
         n = size
-        p = 1.0 / size
-
-        # Random graph
-        G = nx.erdos_renyi_graph(n, p)
-
-        # Adjancency
-        A = nx.to_numpy_array(G)  # shape (n,n), 0/1 entries
-
-        # Center & scale to get the semicircle
-        A_c = (A - p) / numpy.sqrt(n * p * (1-p))
+        X = numpy.random.randn(n, n)
+        X = (numpy.triu(X, 0) + numpy.triu(X, 1).T)
 
-        return A_c
+        return X * (self.r / (2.0 * numpy.sqrt(n)))

{freealg-0.1.7 → freealg-0.1.8}/freealg.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: freealg
-Version: 0.1.7
+Version: 0.1.8
 Summary: Free probability for large matrices
 Keywords: leaderboard bot chat
 Platform: Linux
@@ -30,7 +30,6 @@ Requires-Dist: scipy
 Requires-Dist: texplot
 Requires-Dist: matplotlib
 Requires-Dist: colorcet
-Requires-Dist: networkx
 Requires-Dist: statsmodels
 Provides-Extra: test
 Requires-Dist: tox; extra == "test"

{freealg-0.1.7 → freealg-0.1.8}/freealg.egg-info/requires.txt

@@ -3,7 +3,6 @@ scipy
 texplot
 matplotlib
 colorcet
-networkx
 statsmodels
 
 [docs]

{freealg-0.1.7 → freealg-0.1.8}/requirements.txt

@@ -3,5 +3,4 @@ scipy
 texplot
 matplotlib
 colorcet
-networkx
 statsmodels

freealg-0.1.7/freealg/__version__.py

@@ -1 +0,0 @@
-__version__ = "0.1.7"