Skip to content

Latest commit

 

History

History
45 lines (42 loc) · 3.34 KB

proposal.md

File metadata and controls

45 lines (42 loc) · 3.34 KB
Raw

Project plan/proposal

Below is a general structure, and please refer to the moodle page for your project module for specific format/requirements. If there is any difference, always follow the instructions from the module page.

It is not always a marked component, but very important to ensure a successful project. Think of it as a tool to hellp you think ahead and plan. Also, you will reuse most of it for the interim and final report later.

  • Introduction/background: the general problem area, such as using machine learning to support sensemaking
  • The problem: the specific problem that you want to improve. For example, the exsting recommendations made by machine learning models are not very good.
    • who are your targeted users? and what is the target use case (i.e., what task will they use your tool for)?
    • literature review (academic publications): what academics have done on this problem. More details in the literature review section below.
    • related software, app, website, and other resources: some of the related efforts are not published in an academic paper. For example, not all the recommendaiton algorithms are discussed in a paper.
    • The main goal here is to set your work in the context of existing works: what is the gap/missing piece.
  • Your project idea:
    • The main hypothesis, e.g., using large language models can improve the quality of recommendations
      • This is the idea of how to improve the existing solution.
    • The approach, i.e., the steps to be taken to realise the idea. For example,
      • what large language model do you plan to use? why do you choose it?
      • how do you obtain the model? do you have the hardware needed to run the model?
      • will you need to retrain/update the model? do you have the labeled data for this? do you have hardware required for this?
      • ...
    • The evaluation (following the target paper):
      • what is the definition of 'quality' and how do you measure it?
      • do you plan to talk to/interview your users to understand their requirements?
      • do you plan to do any user evaluations? Qualitative vs. quantitative.
      • ...
  • Plan
    • Set a deadline for each of the task below
      • Make an estimate of how long a task will take, and times that by 4, because you will be doing 3 other modules;
      • Do not have overlapping tasks; this only gives you a false sense of time that you don't actually have;
    1. Minimal working demo (this should be completed before the proposal deadline)
    2. Literature review (if there is more to do)
    3. User requirements:
      1. Talk to the real users
      2. Document their needs: task specific (e.g. finding the most suitable master degree), not technical specific (need to use machine learning)
    4. Paper prototype: design sketches
      1. on real paper or design software such as figma
      2. Get user feedback
    5. Working prototype with minimal functions (MVP: minimal viable product):
      1. leave out non-essentail functions (such as user account) and
      2. use simple alternatives instead of complex ones (e.g., using decisoin tree insteal of Large Language Models)
      3. Get user feedback
    6. Fully working prototype (final prototype)
    7. User evaluation
    8. Final report
  • Create a project plan using GitHub Project (more on this later).