編集履歴

質問編集履歴

読みやすくしました。

2021/05/27 11:22

投稿

valley

スコア0

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -92,7 +92,9 @@
-------------------------------------------------------------------------------------------------------
+----------------------------------------------------------------------------------------------------
+```python
 #!/usr/bin/env python3
@@ -455,3 +457,5 @@
 #
 plt.show()
+```

report7.py追加しました。お願いします

2021/05/27 11:22

投稿

valley

スコア0

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -89,3 +89,369 @@
     renderer.draw_markers(
 TypeError: must be real number, not str
+------------------------------------------------------------------------------------------------------
+#!/usr/bin/env python3
+#coding:utf-8
+#
+# パターンが2次元連続変数で
+# 多変量正規分布(Multivariate Normal Distribution) でモデル化した場合の
+# 識別境界面
+#
+from mpl_toolkits.mplot3d import Axes3D
+import numpy as np
+import matplotlib.pyplot as plt
+import matplotlib.tri as tri
+import matplotlib.cm as cm
+import functools
+import scipy.stats as st
+import scipy.special as sp
+import math
+import sys
+from mpl_toolkits.axes_grid1 import make_axes_locatable
+class Norm2D:
+    """
+    2変量正規分布 (2 dimensional normal distribution)
+    """
+    def __init__(self, X):
+        self.__X     = np.copy(X)
+        self.__Mu    = np.mean(self.__X, axis=0)
+        self.__Sigma = np.cov(self.__X, rowvar=False, bias=True)
+        self.__dim   = len(self.__Mu)
+        self.__N     = len(self.__X)
+    def get_N(self):
+        """ 学習データ数の取得 """
+        return self.__N
+    def get_X(self):
+        """ 学習データの取得 """
+        return self.__X
+    def get_param(self):
+        """ 正規分布パラメタの取得 """
+        return self.__Mu, self.__Sigma
+    def pdf(self, x):
+        """ 与えられた分布の(x)における確率密度値を求める"""
+        return st.multivariate_normal.pdf(x, mean=self.__Mu, cov=self.__Sigma)
+    def sampling(self, N):
+        """ 与えられた分布に従ってN点サンプリングする"""
+        return st.multivariate_normal.rvs(mean=self.__Mu, cov=self.__Sigma, size=N)
+    def this_likelihood(self, x):
+        """ パターンxが与えられた時の与えられた分布での尤度を求める """
+        L = 1.0
+        for n in range(len(x)):
+            L *= self.pdf(x[n])
+        return L
+    def this_log_likelihood(self, x):
+        """ パターンxが与えられた時の与えられた分布での対数尤度を求める """
+        logL = 0.0
+        for n in range(len(x)):
+            logL += log(self.pdf(x[n]))
+        return logL
+# 例題2
+# x(ω1) = {(0, 4√2), (4, 4√2), (2, 4√2-1), (2, 4√2+1)}
+# x(ω2) = {(0, 0), (4, 0), (2, -1), (2, 1), (2(1-√3), 0), (2(1+√3), 0), (2, -√3), (2, √3)}
+"""
+x1 = np.array([[0.0, 4.0*np.sqrt(2.0)],
+               [4.0, 4.0*np.sqrt(2.0)],
+               [2.0, 4.0*np.sqrt(2.0)-1.0],
+               [2.0, 4.0*np.sqrt(2.0)+1.0]])
+x2 = np.array([[0.0, 0.0],
+               [4.0, 0.0],
+               [2.0,-1.0],
+               [2.0, 1.0],
+               [2.0*(1.0-np.sqrt(3.0)), 0],
+               [2.0*(1.0+np.sqrt(3.0)), 0],
+               [2.0,-np.sqrt(3.0)],
+               [2.0, np.sqrt(3.0)]])
+# Set the drow range
+x = y = np.arange(-2, 18, 0.1)
+"""
+#
+# 例題3
+# x(ω1) = {(-2, 0), (0, 1), (0, -1), (2, 0)}
+# x(ω2) = {(3, 2), (4, -1), (6, 1), (7, -2)}
+x1 = np.array([[-2.0, 0.0],
+               [0.0, 1.0],
+               [0.0,-1.0],
+               [2.0, 0.0]])
+x2 = np.array([[3.0, 2.0],
+               [4.0,-1.0],
+               [6.0, 1.0],
+               [7.0,-2.0]])
+# Set the drow range
+x = y = np.arange(-5, 10, 0.1)
+#
+# Create Distributions
+#
+dist1 = Norm2D(x1)
+dist2 = Norm2D(x2)
+#
+# Show Mu and Sigma
+#
+Mu1, Sigma1 = dist1.get_param()
+print('Mu1=', Mu1);
+print('Sigma1=', Sigma1)
+Mu2, Sigma2 = dist2.get_param()
+print('Mu2=', Mu2)
+print('Sigma2=', Sigma2)
+#
+# Create the drawing mesh
+#
+X, Y = np.meshgrid(x, y)
+pos = np.dstack((X,Y))
+#
+# Calc. pdf
+#
+Z1 = dist1.pdf(pos)
+Z2 = dist2.pdf(pos)
+Z = np.log(np.fmax(Z1, Z2))    # Logarithmic expression
+Zdiff = Z1 - Z2
+maxZ = np.max(Z)
+minZ = np.min(Z)
+#print('max Z=', maxZ)
+#print('min Z=', minZ)
+minZ = -20    # Clip the lowest Z value
+#
+# Init. a graph
+#
+fig = plt.figure(figsize=(7,5))
+ax = fig.add_subplot(111, aspect='equal')
+ax.grid()
+plt.xlabel('x')
+plt.ylabel('y')
+#
+# draw the contour
+#
+levels = np.linspace(minZ, maxZ, 50)
+cs = ax.contourf(X, Y, Z, levels=levels, cmap=cm.inferno, extend='both')
+#
+# draw the boundary
+#
+ax.contour(X, Y, Zdiff, levels=[-1.0e-300, 1.0e-300], colors='r', linestyles='-')
+#
+# data
+#
+ax.scatter(dist1.get_X()[:,0], dist1.get_X()[:,1], s=40, c='red',  marker='o', alpha=0.5, linewidths='2')
+ax.scatter(dist2.get_X()[:,0], dist2.get_X()[:,1], s=40, c='blue', marker='o', alpha=0.5, linewidths='2')
+#
+# draw a colorbar
+#
+divider = make_axes_locatable(ax)                       # create the same size divider as the contour graph
+cax = divider.append_axes('right', size='5%', pad=0.1)  # append 5% space to the right-end
+fig.colorbar(cs, cax=cax, format='%.1f')                # show a colorbar
+#
+# Draw
+#
+plt.show()