Merge pull request #35 from Vayun-Goel/linear_regression_2d

Linear regression 2d

ML_Algorithms/Linear_Regression/linear_regression_2d.py

@@ -0,0 +1,62 @@
+import math
+import numpy as np
+
+class LinearRegression2D():
+
+    def __init__(self,points,per_graph_epochs):
+        self.points=points
+        self.per_graph_epochs = per_graph_epochs
+
+    def dot_product(self,x,w,b):
+        z = 0.0
+        z = z + float(x[0])*float(w[0])
+        z = z + float(x[1])*float(w[1]) 
+        z = z + float(b)
+        return float(z)
+
+    def loss_function(self,w,b):
+        mse = 0.0
+        for i in range(len(self.points)):
+            z = self.dot_product(self.points[i],w,b)
+            mse = mse + math.pow((self.points[i][0]-z),2)
+        mse = (mse / float(len(self.points)))
+        return mse
+
+    def grad_descent(self,epochs,w,b):
+        for i in range(epochs):
+            del_w = [0.0 , 0.0]
+            del_b = 0.0
+            m = len(self.points)
+            for j in range(m):
+                h = self.dot_product(self.points[j],w,b)
+                del_w[0] = del_w[0] + (float(self.points[j][0])*(float(h) - float(self.points[j][2])))
+                del_w[1] = del_w[1] + (float(self.points[j][1])*(float(h) - float(self.points[j][2])))
+                del_b = del_b + (float(h) - float(self.points[j][2]))
+            del_w[0] = del_w[0] / float(m)
+            del_w[1] = del_w[1] / float(m)
+            del_b = del_b / float(m)
+            print(del_w)
+            print(del_b)
+            w[0] = w[0] - 0.005*float(del_w[0])
+            w[1] = w[1] - 0.005*float(del_w[1])
+            b = b - 0.05*float(del_b)
+        return w, b
+
+    def visualizer(self):
+        w = [np.random.randn() for i in range(len(self.points[0])-1)]
+        b = np.random.randn()
+        # print(w,b)
+        arr = [[0 for i in range(len(w)+1)] for j in range(7)]
+        for i in range(len(w)):
+            arr[0][i] = w[i]
+        arr[0][2] = b
+        for i in range(5):
+            w,b = self.grad_descent(int(self.per_graph_epochs),w,float(b))
+            arr[i+1][0] = w[0]
+            arr[i+1][1] = w[1]
+            arr[i+1][2] = b
+        w,b = self.grad_descent(2000,w,b)
+        arr[6][0] = w[0]
+        arr[6][1] = w[1]
+        arr[6][2] = b
+        return arr