Feat: Implement Naive Bayes Classifier for MNIST digits.

 On branch master
 Changes to be committed:
	new file:   naive_bayes.py

naive_bayes.py

@@ -0,0 +1,75 @@
+from mxnet import nd, gluon
+import matplotlib.pyplot as plt
+import random
+
+def show_images(imgs, rows, cols):
+ _, axes = plt.subplots(rows, cols)
+ axes = axes.flatten()
+ for i, (ax, img) in enumerate(zip(axes, imgs)):
+ ax.imshow(img)
+ ax.get_xaxis().set_visible(False)
+ ax.get_yaxis().set_visible(False)
+ plt.show()
+ return axes
+
+def transform(data, label):
+ return (data/128).astype('float32').squeeze(axis=-1), label
+
+def train(train, n_classes):
+ X, Y = train[:]
+ n_y = nd.zeros(n_classes)
+ for y in range(n_classes):
+ n_y[y] = (Y==y).sum()
+ P_y = n_y / n_y.sum()
+
+ n_x = nd.zeros((n_classes, 28, 28))
+ for y in range(n_classes):
+ n_x[y] = nd.array(X.asnumpy()[Y==y].sum(axis=0))
+ P_xy = (n_x+1) / (n_y+1).reshape((10, 1, 1))
+ show_images(P_xy.asnumpy(), 2, 5)
+
+ return P_xy, P_y
+
+def predict(img, P_xy, P_y):
+ img = img.expand_dims(axis=0)
+ log_P_xy = nd.log(P_xy)
+ neg_log_P_xy = nd.log(1-P_xy)
+ pxy = log_P_xy * img + neg_log_P_xy * (1-img)
+ pxy = pxy.reshape((10, -1)).sum(axis=1)
+
+ probs = pxy+nd.log(P_y)
+
+ return probs.argmax(axis=0).asscalar()
+
+def test(test, n_classes, P_xy, P_y):
+ X, Y = test[:]
+ correct = 0
+ for i, img in enumerate(X):
+ result = predict(img, P_xy, P_y)
+ if result == Y[i]:
+ correct += 1
+ acc = (correct/X.shape[0]) * 100
+ print("Accuracy {}%".format(acc))
+ return acc
+
+
+def main():
+ n_classes = 10
+
+ mnist_train = gluon.data.vision.datasets.MNIST(train=True, transform=transform)
+ mnist_test = gluon.data.vision.datasets.MNIST(train=False, transform=transform)
+
+ P_xy, P_y = train(mnist_train, n_classes)
+
+ acc = test(mnist_test, n_classes, P_xy, P_y)
+
+ test_X, test_Y = mnist_test[:]
+ index = random.randint(0, test_X.shape[0])
+ result = predict(test_X[index], P_xy, P_y)
+
+ print("Predicted value {}".format(result))
+ plt.imshow(test_X[index].asnumpy())
+ plt.show()
+
+if __name__ == "__main__":
+ main()