最近一直在學習python語言，之前一直使用matlab?？戳撕芏嗖┛投紝懥嗽撍惴ǖ拇a，我覺得還不具體，不詳細。對于初學者來說，看懂不容易。寫了一個NMF算法的python程序供大家參考。參照文獻《Algorithms for non-negative matrix factorization》

具體代碼如下：

import numpy as np import torch import random import matplotlib.pyplot as pltdef nmf(X,r,maxiter,minError):# X=U*V'row,col = X.shapeU = np.around(np.array(np.random.rand(row, r)),5)V = np.around(np.array(np.random.rand(col, r)),5)obj = []for iter in range(maxiter):print('-----------------------------')print('開始第',iter,'次迭代')# update UXV = np.dot(X,V)UVV = np.dot(U,np.dot(V.T,V))U = (U*(XV/np.maximum(UVV,1e-10)))# update VXU = np.dot(X.T,U)VUU = np.dot(V,np.dot(V.T,V))V = (V*(XU/np.maximum(VUU,1e-10)))d = np.diag(1/np.maximum(np.sqrt(np.sum(V*V,0)),1e-10))V = np.dot(V,d)temp = X - np.dot(U,np.transpose(V))error = np.sum(temp*temp)print('error:',error)print('第',iter,'次迭代結束')obj.append(error)if error<minError:breakreturn U, V, objif __name__ =="__main__":X = np.random.randn(20, 50)X = np.array(np.abs(X))#print('X:',X)U,V,obj = nmf(X,2,100,0.01)x = range(len(obj))plt.plot(x, obj)plt.show()

總結

以上是生活随笔為你收集整理的NMF算法python源代码的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。