upload similarity.py

csuldw · csuldw · commit fdba09173fb2 · 2016-02-28T20:17:20.000+08:00
diff --git a/utils/README.md b/utils/README.md
@@ -2,6 +2,7 @@
 
 - data_util.py用于加载MNIST数据集
 - test.py 使用KNN算法测试MNIST数据集
+- similarity.py 用于计算相似度的方法
 
 ## 依赖
 
diff --git a/utils/similarity.py b/utils/similarity.py
@@ -0,0 +1,32 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Sun Feb 28 19:38:58 2016
+similarity
+@author: liudiwei
+"""
+
+import numpy as np
+
+#欧几里得距离，使用1/(1+distance)归一化
+def eulisSim(matA, matB):
+    return 1.0/(1.0 + np.linalg.norm(matA - matB))
+
+#pearson相关系数，使用0.5+0.5*corrcof()将数值归一化
+def pearsSim(matA, matB):
+    if len(matA) < 3:
+        return 1.0
+    return 0.5 + 0.5 * np.corrcoef(matA, matB, rowvar = 0)[0][1]
+
+#余弦相似度，使用0.5+0.5*cos(theta)归一化
+def cosSim(matA, matB):
+    val1 = float(matA * matB.T)
+    val2 = np.linalg.norm(matA) * np.linalg.norm(matB)
+    cos_theta = val1/val2
+    return 0.5 + 0.5 * cos_theta
+    
+if __name__=="__main__":
+    matA = np.mat([1,2,3,4,5])
+    matB = np.mat([2,3,4,5,6])
+    edist = eulisSim(matA, matB)  #output: 0.3090169943749474
+    pdist = pearsSim(matA, matB)  #output: 1.0
+    cdist = cosSim(matA, matB)    #output: 0.99746833816309111
