lsa

CH17/lsa.py

@@ -1,15 +1,34 @@
 #! /usr/bin/env python
+
 # Project:  Lihang
 # Filename: lsa
 # Date: 6/03/19
 # Author: 😏 <smirk dot cao at gmail dot com>
 # 截断奇异值分解用在count/tf-idf矩阵的时候，叫做潜在语义分析。
+import numpy as np
 
 
 class LSA(object):
     def __init__(self, n_components):
         self.n_components = n_components
         self.components = None
+        self.singular_values = None
+        self.explained_variance_ratio = None
 
     def fit(self, x):
-        pass
+        u, s, vh = np.linalg.svd(x, full_matrices=False)
+        max_abs_raws = np.argmax(np.abs(vh), axis=1)
+        signs = np.sign(vh[range(vh.shape[0]), max_abs_raws])
+        u *= signs
+        vh *= signs[:, np.newaxis]
+        k = self.n_components
+        # 
+        self.components = vh[:k]
+        self.singular_values = s[:k]
+        # 
+        x_transformed = u*s
+        self.explained_variance = np.var(x_transformed, axis=0)
+        self.explained_variance_ratio = (self.explained_variance/self.explained_variance.sum())[:k]
+        self.explained_variance = self.explained_variance[:k]
+        return x_transformed
+

CH17/unit_test.py

@@ -99,3 +99,34 @@ def test_lsa_1701(self):
         print("singular_values\n", lsa1.singular_values_)
         print("components\n", lsa1.components_)
         print("rst\n", rst)
+
+    def test_lsa(self):
+        x = np.array([[0., 0., 1., 1., 0., 0., 0., 0., 0.],
+                      [0., 0., 0., 0., 0., 1., 0., 0., 1.],
+                      [0., 1., 0., 0., 0., 0., 0., 1., 0.],
+                      [0., 0., 0., 0., 0., 0., 1., 0., 1.],
+                      [1., 0., 0., 0., 0., 1., 0., 0., 0.],
+                      [1., 1., 1., 1., 1., 1., 1., 1., 1.],
+                      [1., 0., 1., 0., 0., 0., 0., 0., 0.],
+                      [0., 0., 0., 0., 0., 0., 1., 0., 1.],
+                      [0., 0., 0., 0., 0., 2., 0., 0., 1.],
+                      [1., 0., 1., 0., 0., 0., 0., 1., 0.],
+                      [0., 0., 0., 1., 1., 0., 0., 0., 0.]])
+
+        svd = lsa_sklearn(n_components=3, n_iter=7, random_state=42)
+        svd.fit(x)
+        print("\n")
+        print("lsa sklearn")
+        print("components_\n", svd.components_)
+        print(svd.singular_values_)
+        print(svd.explained_variance_)
+        print(svd.explained_variance_ratio_)
+        print(svd.explained_variance_ratio_.sum())
+
+        svd_1 = lsa_test(n_components=3)
+        svd_1.fit(x)
+        print("lsa test")
+        print(svd_1.components)
+        print(svd_1.singular_values)
+        print(svd_1.explained_variance)
+        print(svd_1.explained_variance_ratio)