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.