Additional Reading

• Marinai et al: Artificial Neural Networks for Document Analysis and Recognition View
• LeCun et al: Gradient-based learning applied to document recognition View

Exercises

1) Download the following files from attachments:

train-images-idx3-ubyte.gz
train-labels-idx1-ubyte.gz
mnist.py

2) Add following lines of code in your python script to get mnist database in python

from mnist import *
images = mnist_read("train-images-idx3-ubyte.gz")
classes = mnist_read("train-labels-idx1-ubyte.gz")

Write a perceptron classifier to separate “6” and “3”. Divide your data into suitable training and testing sets.
e.g; use training set sizes of 100, 1000, 10000 etc. and discuss your results.