added comments to hw1

perceptron.py

@@ -12,28 +12,39 @@
 y_test = map(lambda x: x[0], data_test)
 xs_test = map(lambda x: (1, x[1], x[2]), data_test)
 
+"""
+p_train takes a list of x vectors and a list of their corresponding classification
+and returns a function representing the learned decision boundary
+"""
 def p_train(xs,ys,shuffled=False,epoch=0):
-	w = list(repeat(0, len(xs[0])))
+    # Create initial vector of zero weights
+    w = list(repeat(0, len(xs[0])))
+    # function used for taking dot product of vectors
 	dot_product = lambda x, w: sum(imap(mul, x, w))
 	updated = True
+    # pack the input vectors with their corresponding outputs
 	data = zip(xs,ys)
-	if shuffled:
-		random.shuffle(data)
-	if epoch:
+    # if we are doing partial epochs slice the list from zero to epoch
+    if epoch:
 		data = data[:epoch]
 	while updated:
+        # if we are shuffling our epochs do so
 		if shuffled:
 			random.shuffle(data)
 		updated = False
 		for x,y in data:
-			print y
+            # check if data is correctly classified
 			if dot_product(x, w) * y <= 0:
+                # if not adjust the weight vector
 				w = [y * xi + wi for xi, wi in izip(x, w)]
 				updated = True
-	return lambda x: -(w[0] + x * w[1]) / w[2]	
+    # return a function creates the learned decision boundary
+	return lambda x: -(w[0] + x * w[1]) / w[2]
 
-x = range(-4,9)
 
+# Graph generation
+
+x = range(-4,9)
 font = {'family' : 'normal',
         'weight' : 'bold',
         'size'   : 14}

regression.py

@@ -1,7 +1,7 @@
 from __future__ import division
 from operator import mul
 from itertools import imap, izip, repeat
-from numpy import matrix 
+from numpy import matrix
 import numpy as np
 
 f = open('tn-train.txt')
@@ -16,7 +16,10 @@
 
 xs = [list(repeat(1,len(x1))) ,x1, x2]
 
-
+"""
+Given a list of input vectors and the corresponding output learn a linear regression
+and return a function which predicts y given 2 inputs
+"""
 def multi_regression(xs,y):
     xs = matrix(xs)
     xTy = np.dot(xs,y)
@@ -27,7 +30,9 @@ def multi_regression(xs,y):
     print type(w[0])
     return lambda x1,x2 : w[0] + x1*w[1] + x2*w[2]
 
-
+"""
+Given a list of x and y values return a function which predicts y for a given x
+"""
 def linear_regression(x, y):
     avg = lambda data: sum(data) / len(data)
     xy_bar = avg(map(mul, x, y))
@@ -38,6 +43,15 @@ def linear_regression(x, y):
     b = y_bar - w * x_bar
     return lambda x: w*x + b
 
+
+"""
+Tests and SSE
+
+In our document we misunderstood the SSE who our values are too low as we divided by the number of tests
+so the SSE is off by a factor of the len(y)
+"""
+
+
 f = multi_regression(xs,y)
 sse = 0
 for xi,xii,output in zip(x1,x2,y):