What is this template?
A simple repo with a single README meant to serve as a template for starting research projects. Use this template and fill in the questions below! Primarily aimed at CS research but apply to your own domain as needed ๐.
The original idea was taken from this blog post.
Some more advice before you get started:
- A 4-step system for your research paper strategy
- 7 Time Management Tips for Academic Writers to Avoid Crunch Time before an Academic Deadline
- Use the PEEL Technique to Structure your Paragraph Writing
Research Outline:
Use this section to set an outline and guiding vision describing your work.
The Problem (What)
- A brief description of the problem we're hoping to solve with our work.
Why even do this? (Why)
- An overview of why this would matter and why it is worth solving.
How do current approaches fail or fall short?
- Come up with a list of current or related approaches and why they either won't work or underperform.
- This also acts as a small lit review & can use to set up baselines for comparisons.
What does our solution look like?
- Map what the solution is expcted to look like. This will likely change over the course of the project.
When is the problem considered "solved"?
- A general idea of what evaluation metrics will be used and in some cases what is the quantity required for it to be "solved".
What are potential pitfalls or known unknowns?
- Outline what might cause the project to fail, whether that is incomplete knowledge, known shortcomings of your approach, or areas of high uncertainity.
What is the plan?
- Create a list of steps and action items needed to get started working.
- Consider using project management tools like Github's Issues or Project pages to help keep track of work.
Documentation
Use this section to get a head start in ensuring your work is reproducible and readers hoping to apply your research can follow along with little hassle.
Setup
- Describe how to set up the project. What are the dependencies, any hardware/software requirements, where/how can we download the data used?
Train / Test
- If you created an ML system, how can we run the training loop on our machine? How do evaluate the model the same way as the paper?
- If this isn't an ML system you can still describe the testing/eval process here so others can recreate your work easily.
Usage
- Share example usage code or describe how someone might apply your solution to their own domain.