In this project, you will evaluate CNNs, RNNs, and potentially combinations of CNN+RNN architectures. You may also explore neural network topics beyond what we cover in class. These include, but are not limited to:
- Generative Models (e.g., GANs, VAEs)
- Transformer Architectures
- Deep Reinforcement Learning
To encourage exploration, evaluating an RNN or another post-CNN algorithm is mandatory. Failure to do so will result in a loss of creativity points as outlined in the rubric at the end of this document.
The goal is to:
- Optimize performance on a particular task, or
- Apply a technique to a research area of interest.
You may use any tools to implement these networks, such as PyTorch, Keras, TensorFlow, etc. The assumption is that you already understand the principles of training, backpropagation, and optimization, and are ready to use these frameworks that handle many of these operations for you.
You may use the datasets provided below or choose a project related to your own research/interests. If you're pursuing a custom project:
- Approval is required. Please email Prof. Kao directly to get approval.
- Projects must explore a post-CNN topic to be considered.
- Custom project proposals must be submitted by Friday, 1 March 2024. No approval will be granted after this date.
You may work individually or in groups of up to 4 people. To find teammates, use the "Search for Teammates" functionality on Piazza: Search for Teammates.