CSSE 413: Take-home Final

• The take-home final is due Thursday, May 23rd at noon. I imagine that for each question, you write about two paragraphs to half a page.
• [8 pts] Explain the architecture and training regimen of a feed-forward network.
• [12 pts] Explain the architecture and training regimen of a convolutional neural network.
• [20 pts] Explain the architecture and training regimen used for a transformer based auto-regressive NLP system such as ChatGPT.
• [20 pts] Explain how the size of a network as well as key properties of the data set used for training affects the quality and performance of neural network systems.
• [10 pts] Explain ethical issues surrounding copyright of and bias in the data.
• [20 points] Explain pattern recognition, feature learning, knowledge representation and reasoning in key neural network architectures.
• [10 pts] Explain how value iteration and policy iteration work. Explain how Q-Learning differs from the prior two algorithms.
• Submit a pdf copy of your work which includes your name near the top of the document, to the appropriate drop-box on Moodle.