The Value of Group Projects for PhD Students

Tim Kovacs. 2007.

I think a lot of research experience can be had from group projects for relatively little effort. Here are my notes from a presentation to the Machine Learning and Biological Computation group on the subject.

Observations on the MLBC group

The group has a lot of expertise in different areas:

• Research skills (planning, reviewing, writing, presenting...)
• Many areas of machine learning, maths, logic...
• Tools: weka, eclipse, R, cvs, Prolog, LaTeX...

Problem: Interaction between members is limited

• We work in diverse areas
• There's nothing driving collaboration
• We're all busy

Solution:

• Find areas of common interest
• Collaborate on group projects

PhD Study

Problems:

• Research skills are hard to learn and hard to teach
• Best learned by doing...but doing a PhD is slow
• Can feel isolated, overwhelmed by size of project

Solutions:

• Smaller, faster research projects
• Group projects
  • Learn from others
  • Less time-consuming than doing it all yourself
• Alternative: supervise final year/MSc projects

A proposal

• We select a project
• Participation is optional
• No minimum commitment (but do follow through on what you do commit to)
• Weekly meetings
  • Can replace some regular group meetings
• Volunteer for tasks:
  • Brainstorming
  • Literature searches/reviews
  • Giving tutorials and presentations to group, or updating rest of research group
  • Implementing
  • Experimenting
  • Writing-up
  • Coordinating project

Selecting a project

Projects should:

• Have broad interest
• Be fun e.g. robotics, games...
• Be incremental and parallelisable
• Have concrete goals e.g.
  • Submit to a particular workshop
  • Enter a competition
  • Improve on an existing algorithm
  • Produce some useful software
  • Demonstrate a proof of concept

---