AI-Assisted Research and Coding

Modern AI tools may be used throughout the course to support literature exploration, concept development, coding, debugging, and the creation of visual explanations. Week 2 will provide a hands-on introduction to agent-based research and coding workflows.

These tools should extend—not replace—your own scientific reasoning. You remain responsible for:

• verifying technical claims against primary sources;
• testing and understanding submitted code;
• checking results using appropriate numerical and physical reasoning;
• citing sources and acknowledging substantial AI assistance; and
• ensuring that all submitted work reflects your own understanding and judgment.

The quality of your work will be assessed through its scientific insight, clarity, correctness, and evidence—not through whether AI tools were used.

Coursework & Grading

Paper Presentation and Discussion Participation — 25%

Paper presentation: 20%. Teams will present two related papers and lead the subsequent class discussion. Presentations should build intuition, connect the papers, and use effective visual explanations rather than simply reproducing their contents.

Discussion participation: 5%. Non-presenting students will sign up for papers and work in small groups during the first 10 minutes of class. Groups will record their observations and questions, then selected group leads will contribute to the discussion after the presentation.

Interactive Lab — 25%

Teams of two or three will develop three connected notebooks exploring a course topic: a toy example, a paper-based investigation, and an extension. During class, another group will explore the notebooks and then deliver a short reverse pitch presenting the lab's key ideas, results, and what they learned through interacting with it.

Project Proposal Pitch — 10%

Students may work individually or in pairs. The pitch should define the research goal, position it relative to prior work, summarize relevant methods, and present an initial plan or, optionally, preliminary results.

Final Project Report — 40%

The final project is an open-ended physics-informed machine-learning investigation. The report should follow the CVPR format and be 6–8 pages, including references, with an absolute maximum of 8 pages.

Schedule (Fall 2026)

On student-presentation weeks, Group A leads the first hour and Group B leads the second. See the full schedule and reading list for paper links, background readings, and discussion materials.

Week	Date	Topic	Format and readings
1	Sep 9	Course overview and introduction	Instructor-led
2	Sep 16	Hands-on agent-based research and coding seminar	Led by Rohan Dahale
3	Sep 23	Neural Fields Across Astronomical Scales	Instructor research presentation
4	Sep 30	Neural fields: representations and inverse problems	Group A: Representing high-frequency signals Group B: Neural fields in inverse problems Lab topic declaration: pitch your lab idea and submit the initial README by the end of Week 4.
5	Oct 7	Conditioned neural fields: shape families, topology, and discontinuities	Group A: Representing families of shapes Group B: Topology and discontinuities through lifting
6	Oct 14	Physics-informed learning: solving and discovering equations	Group A: Known physics, hidden solutions Group B: Unknown physics and hidden coordinates
7	Oct 21	Neural differential equations and structured dynamics	Group A: Neural ordinary differential equations Group B: Learning mechanics with physical structure
—	Oct 28	Reading Week — no class	No class
8	Nov 4	Integrative interactive lab	Notebook exploration and reverse pitches Lab due Nov 3, 23:59
9	Nov 11	Neural operators and parametric solution maps	Group A: Foundational neural operators Group B: Continuous fields and changing geometry
10	Nov 18	Diffusion models for scientific inverse problems	Group A: From pixel-space to latent diffusion Group B: Generative priors without complete observations
11	Nov 25	Project proposal presentations I	Individual students or pairs
12	Dec 2	Project proposal presentations II	Final class
—	Dec 8	End of term	Final project due Dec 15

Policies

Academic Integrity and Attribution

Submitted work must reflect your own understanding and scientific judgment. Cite papers, software, datasets, code, figures, and other sources appropriately, and acknowledge substantial assistance from AI tools. You are responsible for verifying every claim, result, and piece of code included in your work.

Collaboration

Discussion and exchange of ideas are encouraged. For team assignments, all members should understand the complete submission and contribute meaningfully. Work submitted for an individual component must be your own unless collaboration is explicitly permitted.

Deadlines and Accommodations

Assignment deadlines are listed in the course GitHub repository. If illness, accessibility needs, or other circumstances may affect your work, contact the instructor or teaching assistant as early as possible so that appropriate arrangements can be discussed.

Communication and Announcements

Discord is the main channel for coordination and informal discussion; the invitation is available on Quercus. Official announcements, course materials, and grading information will also be posted on Quercus.