CSSE 313: Days 14 and 15

Outline

• Convolutional neural networks (CNNs):
  • Justification
  • How they work
  • Demo of a CNN to recognize digits
  • Pattern recognition in CNNs
  • An early, classic CNN: LeNet5
• Introduction to CNN assignment

Resources

• A visualization of a CNN that recognizes digits
• Pattern recognition in CNNs
• Slides on CNNs
• Softmax
• Softmax in MNIST FFNet
• Wikipedia article on LeNet5
• Slides about Digit Recognition and data curation in LeNet.
• Introduction to CNN assignment: Worksheet about LeNet5 architecture Source: https://en.wikipedia.org/wiki/LeNet#/media/File:LeNet_architecture.png
• Original article about LeNet5

Assignments

• Day 16, BC: Summarize the AlphaGoZero paper and submit your summary to the appropriate drop-box on Moodle. When it comes to the length, 250 - 300 words seems to work well for the purposes of our class. We will discuss this article in class. This is a very technical article. You may have to read it more than once. The article is quite short, but it also has a wealth of information in the appendix. I strongly encourage you to read it.