homework

sherlock-coding · sherlock-coding · commit 09f603ba6d8a · 2014-12-16T14:03:31.000+08:00
diff --git a/Homework/HousingPricePrediction.py b/Homework/HousingPricePrediction.py
@@ -0,0 +1,65 @@
+#LMS算法
+import math
+
+#求h(x,theta)
+def GetHThetaX(x,theta):
+        theta_1 = theta[1:]
+        h = theta[0] + sum([xi*thetai for xi,thetai in zip(x,theta_1)])
+        return h
+
+def GetJ(X,Y,theta):
+        J = sum(pow(GetHThetaX(x,theta)-y,2) for x,y in zip(X,Y))/2
+        return J
+        
+
+#求梯度
+#j表示对thetaj求偏导数
+def GetGradient(X,Y,theta,j):
+        h = []
+        for x in X:
+                h.append(GetHThetaX(x,theta))
+        g = 0;
+        if j==0:
+                g = sum([hi-yi for hi,yi in zip(h,Y)])
+        else:
+                g = sum([(hi-yi)*x[j-1] for hi,yi,x in zip(h,Y,X)])
+        return g
+
+
+#n表示n维的特征向量
+def LMS_Prediction(X,Y,n,year):
+        #theta全都初始化为0
+        theta = [i for i in range(0,n)]
+
+        #learning rate设置
+        alpha = 0.00000001
+
+        steps = 0
+
+        curJ = GetJ(X,Y,theta)
+        preJ = curJ
+                
+        while True:
+                temp = []
+                for i in range(0,n):
+                        g = GetGradient(X,Y,theta,i)
+                        #print(g)
+                        temp.append(theta[i]-alpha*g)
+                #print(theta)
+                theta = temp
+                print(steps,":",temp)
+                steps += 1
+                preJ = curJ
+                curJ = GetJ(X,Y,theta)
+                if curJ>preJ:
+                        break
+        
+        
+
+if __name__ == '__main__':
+        #X = [[1],[2],[5],[7], [10], [15]]
+        #Y = [2, 6, 7, 9, 14, 19]
+	X = [[2000],[2001],[2002],[2003],[2004],[2005],[2006],[2007],[2008],[2009],[2010],[2011],[2012],[2013]]
+	Y = [2.000,2.500,2.900,3.147,4.515,4.903,5.365,5.704,6.853,7.971,8.561,10.000,11.280,12.900]
+	LMS_Prediction(X,Y,2,2014)
+	
