Students are required to conduct and submit a research project as well as present an accompanying poster at the poster session on March 12. Students may work on projects in groups of up to 3 people. Note the topic of your project can be different from the topic of your presentation.
The project can be on any topic related to the course material. The purpose of the project is to gain a better understanding of experimental and/or theoretical aspects of interactive learning algorithms. While we encourage you to explore new research directions, projects may also focus on reproducing or comparing previous approaches, perhaps investigating various theoretical or design choices. If you are a graduate student currently working on a related topic, you may use your research for the project. Generally, projects will fall into one of the following categories:
Pure Theory: study proof techniques, try to extend a proof, or apply a proof to a new setting.
Algorithms & Models: extend an algorithm or model, design a new one, or adapt one for a new setting.
Applications: identify the correct assumptions for approaching a new setting, experimentally validate these assumptions.
Due Feb 2nd.
The project proposal is a mechanism to get you to start thinking about your project. It is non-binding; you can change the direction of the project after the proposal submission.
Create a folder in the Google drive (accessible via your UChicago email address) and upload your project proposal.
Project proposals could be 2-3 pages (3 pages maximum), excluding references.
The structure for the project proposal should generally cover the following aspects:
Introduction of research area: Motivate why this area is interesting and describe key technical challenges.
Related work: Give an overview of different types of approaches and highlight assumptions of previous work. This should not be a laundry list of previous papers. Rather, it should describe previous approaches in broad strokes.
Research question(s): Describe one or more research questions to be addressed. The research question should be a high-level description of the project topic.
Research plan: Describe a preliminary plan to address the research question(s), and outline any experimental set-ups, such as datasets, models, etc. The research plan should extend the research question(s) to the details of the project.
The project can be on any topic related to the topics covered in the lecture . You are encourage to combine the course project with your own research domain. Possible projects include, but not limited to, one (or more) of the following (interleaving) directions:
Propose an active learning algorithm on a real-world ML scenario where data resource is scarce. Understand what are the challenges (e.g., constraints, structure of the data) posed by the problem domain; explore how you can leverage these knowledge or adapt to the these constraints when designing the active learning algorithm; evaluate your algorithm to see whether there is an improvement over traditional supervised approaches.
Explore new data domains which renders supervised learning approaches inappropriate. For instance, one practical scenario is streaming data, where the learning/decision making system is scanning through a large volume of data sequentially and only visit each data point (at most) once, while (sequentially) making decisions for each incoming data point . Possibly applications scenario include data in the physical sciences (high-energy physics), financial domains (high-frequency trading), multimedia (massive data summarization on the fly), etc.
Extend existing advanced supervised models to the adaptive/interactive setting, and develop a theoretical understanding in why it works, and evaluate your approach on synthetic or real-world sequential decision problems that are suitable for you model. For example, "activise" a (e.g., Bayesian) deep neural net for experimental design; adapt advanced supervised learning approaches in the context of reinforcement learning or imitation learning applications.
Explore novel learning protocols for interactive learning systems, and proposal learning algorithms that are tailored to the rich data (e.g., pre-collected dataset, or acquired through your interface). For example, develop and evaluate novel active learning algorithms that could utilize multiple sources of information, and/or explore the trade-offs in the benefit and additional cost introduced by such rich sources of information.
Design novel interfaces for learning systems that interact with the real-life environment. For example, when interacting with a human agent, design and implement an interface that are natural for a human learner to perceive; understand what are the key factors that should be considered based on the nature of the environment.
Develop novel machine teaching algorithms for specific types of learner model, or for new machine teaching application. For example, teaching a "simulated" learner represented by a modern machine learning algorithm; teaching an interactive machine learning agent (e.g., teaching a robot towards approaching a specific goal). Evaluate your machine teaching algorithm under the corresponding task scenario.
Information on poster and project report coming soon.