Differences

This shows you the differences between two versions of the page.

--- code:perceptron [2013/08/15 10:42]
asa created
+++ code:perceptron [2016/08/09 10:25]
127.0.0.1 external edit
@@ Line 7: / Line 7: @@
 import numpy as np
-from PyML.classifiers.baseClassifiers import Classifier
+from matplotlib import pyplot as plt
-class Perceptron (Classifier) :
+class Perceptron :
-    """An implementation of the perceptron algorithm"""
+    """An implementation of the perceptron algorithm.
+    Note that this implementation does not include a bias term"""
-    attributes = {'max_iterations': 50,
+    def __init__(self, max_iterations=100, learning_rate=0.2) :
-                  'learning_rate' : 0.2}
-    def __init__(self, arg = None, **args) :
+        self.max_iterations = max_iterations
+        self.learning_rate = learning_rate
-        Classifier.__init__(self, arg, **args)
+    def fit(self, X, y) :
+        """
+        Train a classifier using the perceptron training algorithm.
+        After training the attribute 'w' will contain the perceptron weight vector.
+        Parameters
+        ----------
-    def train(self, data, **args) :
+        X : ndarray, shape (num_examples, n_features)
+        Training data.
-        Classifier.train(self, data, **args)
+        y : ndarray, shape (n_examples,)
+        Array of labels.
-        if data.labels.numClasses != 2 :
-            raise ValueError, "not a binary classification problem"
+        """
+        self.w = np.zeros(len(X[0]))
-        self.bias = 0
-        data.addFeature('bias', [1.0 for i in range(len(data))])
-        num_features = data.numFeatures
-        self.w = np.zeros(data.numFeatures)
         converged = False
         iterations = 0
         while (not converged and iterations < self.max_iterations) :
             converged = True
-            for i in range(len(data)) :
+            for i in range(len(X)) :
-                label = 2 * data.labels.Y[i] -1 # convert the labels from 0,1 to -1, +1
+                if y[i] * self.discriminant(X[i]) <= 0 :
-                if label * self.decisionFunc(data, i) <= 0 :
+                    self.w = self.w + y[i] * self.learning_rate * X[i]
-                    self.w = self.w + label * self.learning_rate * data.X[i]
                     converged = False
+                    plot_data(X, y, self.w)
             iterations += 1
-        data.eliminateFeatures([data.numFeatures -1])
-        self.bias = self.w[-1]
-        self.w = self.w[:-1]
         self.converged = converged
         if converged :
-            print 'converged in %d iterations ', iterations
+            print 'converged in %d iterations ' % iterations
-        # this should be the last command in the train function
-        self.log.trainingTime = self.getTrainingTime()
+    def discriminant(self, x) :
+        return np.dot(self.w, x)
+    def predict(self, X) :
+        """
+        make predictions using a trained linear classifier
+        Parameters
+        ----------
-    def decisionFunc(self, data, i) :
+        X : ndarray, shape (num_examples, n_features)
+        Training data.
+        """
+        scores = np.dot(self.w, X)
+        return np.sign(scores)
-        return np.dot(self.w, data.X[i]) + self.bias
+def generate_separable_data(N) :
+    xA,yA,xB,yB = [np.random.uniform(-1, 1) for i in range(4)]
+    w = np.random.uniform(-1, 1, 2)
+    print w,w.shape
+    X = np.random.uniform(-1, 1, [N, 2])
+    print X,X.shape
+    y = np.sign(np.dot(X, w))
+    return X,y,w
+def plot_data(X, y, w) :
+    fig = plt.figure(figsize=(5,5))
+    plt.xlim(-1,1)
+    plt.ylim(-1,1)
+    a = -w[0]/w[1]
+    pts = np.linspace(-1,1)
+    plt.plot(pts, a*pts, 'k-')
+    cols = {1: 'r', -1: 'b'}
+    for i in range(len(X)):
+        plt.plot(X[i][0], X[i][1], cols[y[i]]+'o')
+    plt.show()
-    def classify(self, data, i) :
+if __name__=='__main__' :
+    X,y,w = generate_separable_data(40)
+    p = Perceptron()
+    p.fit(X,y)
-        score = self.decisionFunc(data, i)
-        classification = 1 if score > 0 else 0
-        return (classification, score)
 </file>

CS545 fall 2016

User Tools

Site Tools

Differences

Page Tools