# Project Milestone 3: Abstract + Completion Plan

**Due: March 6 (submit PDF on Blackboard)**

This milestone has two parts: a formal project abstract and a week-by-week completion plan.

By this point you've had time to think about two directions. Now it's time to commit to one — or a hybrid — and sharpen it into a concrete research plan. The goal isn't to have everything figured out, but to have a clear enough direction that you can make steady progress through the rest of the semester.

---

## Part 1: Abstract

Write a formal abstract for your project (roughly 200–300 words). This is the kind of abstract you'd write for a conference paper submission — concise, specific, and self-contained.

Your abstract should address all of the following:

### Working title

Provide a working title. It doesn't have to be final, but it should reflect what your project is actually about.

### Research question and hypothesis

This is the most important part. By M3, your research question should be specific enough to generate a **testable scientific hypothesis** — a quantifiable prediction that your experiments can support or refute.

A hypothesis is stronger than a question: it commits to an expected direction or magnitude. For example:
- *Question: Does self-supervised pretraining help when labeled data is limited?*
- *Hypothesis: Self-supervised pretraining on unlabeled domain data will improve classification accuracy by at least X% over training from scratch, given only N labeled examples.*

You don't need to be right — that's the point of running the experiment. But you need to make a prediction, not just observe what happens.

### Data

Name the specific dataset(s) you will use and provide a citation or URL. If you plan to collect or compile data yourself, describe the source and scope concretely.

Data access is often the biggest source of project delays. If there's any uncertainty about whether you can actually obtain the data, address it here. If you still have concerns, flag them in Part 2 (ask for guidance).

### Methods

Describe the data-efficient ML technique(s) you will apply and the baseline you'll compare against. What are you actually doing, in concrete terms?

### Expected findings

What do you expect to find, and why? This doesn't have to be elaborate — a sentence or two connecting your hypothesis to the methods is enough.

### Significance

Briefly: why does this matter? What larger question does it connect to?

---

## Part 2: Completion Plan

List all broad tasks needed to complete the project, organized week by week from now through the end of the semester. Your plan should account for all remaining milestones and deadlines:

| Date | Milestone |
|------|-----------|
| Mar 24 | M4: Progress report + GitHub repo |
| Apr 7 | Code walkthrough (30-min meeting) |
| Apr 14 | M5: Final progress report |
| Apr 28–30 | Presentations |
| Finals week | Final writeup due |

A good completion plan:
- Breaks the project into concrete tasks (not just "work on model")
- Is realistic about what's achievable in each window — spring break is March 7–15
- Shows a coherent arc from data → code → experiments → analysis → writing

If you're working in a group, specify which tasks each person is responsible for and briefly explain why the division makes sense given your backgrounds and interests.

---

## Ask for Guidance

Use this milestone to surface anything you'd like help with. I can respond in writing, or we can meet to talk through it.

Some things worth flagging if they apply to you:
- Uncertainty about data access or feasibility
- Modeling decisions you're not sure how to make
- Questions about scope (too ambitious? not ambitious enough?)
- Anything about the experimental design that feels underspecified

---

## Formatting and logistics

- Submit as a single PDF on Blackboard
- No strict length requirement; 2–3 pages is typical
- If working in a group, all members submit the same document