Description
1. Design neural networks to classify hand written numbers using keras library, however
WITHOUT including convolutional layers.
(a) Design neural networks and implement them.
• Find Assignment 6 mnist.ipynb and mnist.zip.
• Unzip mnist.zip and locate the files in ‘mnist’ folder.
• train X and test X are list of intensities in 8-bit gray scale of 28 × 28 images.
• train y and test y are list of integers which are the classes of the corresponding images.
• The objective of the networks is classifying images into 0 to 9.
(b) (30 points) Report the top three best network designs and their test accuracies
separately as Figure 1. Submit html and ipynb files which include your source
codes and pdf file which includes your report.
(c) (Extra points) Based on the best test accuracy of each group, Rank #1 group will
receive extra 5 points and Rank #2 group will receive extra 3 points.
Figure 1: A network design example