In this course, you will learn advanced topics in Generative AI. Part of the learning will be through in-class lectures and take-home assignments, but you will really gain hands-on experience with Generative AI in your course project. We would like you to choose wisely a project that fits your interests. One that would be both motivating and technically challenging.
You will participate in the project in small groups, with a maximum of 5 people. The project contributes to 50% of the course credits:
5 % proposal
15 % milestone report
20 % final report
10 % final presentation
In the end, you will write an 8-page report about your project. The quality of this report should be close to a top machine learning conference workshop paper.
Most students choose one of four types of projects:
Theoretical project. Analyze and prove some interesting properties of a GenAI model or a learning algorithm.
Algorithmic project. Develop a new learning algorithm, or a novel variant of an existing algorithm, for GenAI.
Modeling project. Develop a new deep neural network architecture, or a novel variant of an existing model for GenAI.
Application project. Apply a model and a learning algorithm to solve a novel GenAI application of interests more efficiently.
Many fantastic projects come from students picking either an application that they’re interested in, or picking some subfield of GenAI that they want to explore more. So, pick something that you are passionate about! Be brave rather than timid, and do feel free to propose ambitious things that you’re excited about. ( Just be sure to ask us for help if you’re uncertain how to best get started.) Alternatively, if you’re already working on a research project that GenAI might apply to, then you may already have a great project idea.
Your project will be evaluated based on criteria that are similar to a research paper.
Novelty. Is this project applying a common technique to a well-studied problem, or is the problem or method relatively unexplored?
Significance. Did the authors choose an interesting or a “real” problem to work on, or only a small “toy” problem? Is this work likely to be useful and/or have an impact?
Technical Quality. Does the technical material make sense? Are the things tried reasonable? Are the proposed algorithms or applications clever and interesting? Do the authors convey novel insight about the problem and/or algorithms?
Presentation and Writing. Is the idea and the solution clearly conveyed? Are the figures and tables carefully crafted? Is the report well structured and well reasoned?
Hint: A very good project will be a publishable or nearly-publishable piece of work. Three of the main machine learning conferences are ICML, NeurIPS and ICLR. All papers from these machine learning conferences are available online.
You can browse some of the recent machine learning papers to get inspired.
Data: many benchmark datasets are freely available, e.g. MNIST, CIFAR-10, SQuAD. You can also search public datasets at:
Compute:
UC San Diego provides Data Science/Machine Learning platform for this course. You can log in with your Active Directory ID.
In addition, you can apply for student cloud computing credits at Google Cloud and Amazon AWS.
Image Animation
Image animation translates the motion of a driving video to animate an object in a target image. This project aims to generate a realistic video to animate objects in a diverse set of target images. A popular method in image animation is First Order Motion Model. You can learn more about image animation in this NeurIPS paper.
Automatic Drug Design
Drug discovery finds target molecules with desired chemical properties. Current drug discovery and development pipelines are long, complex and depend on numerous factors. This project aims to speed up drug discovery by automatically generating molecular graphs. The common datasets include ChEMBL, ZINC. You can learn more about this problem and solutions from this ICML paper and this github repo.
AI Data Scientist
An autonomous agent capable of handling the end-to-end data science workflow. The system will proactively identify and acquire relevant datasets from multiple sources, perform data cleaning and preprocessing, and automatically select and fine-tune appropriate models based on the problem context. By integrating natural language interfaces, the agent will allow users to pose research questions in plain English and receive well-structured analyses and visualizations. You can learn more about this problem following AI Scientist V2 and this blog post.
Story Visualization [Final Report]
LLM Pseudo-Random Number Generation [Final Report]
ConceptDiffusion: Controlled Image Generation with Concept-Conditioned Diffusion Models [Final Report]