User:Jadrian Miles/Plan your PhD

This page includes feedback from Jadrian's 2008-12-09 sciviz lunch talk.

• demos are not necessary; just talking to visitors or showing slides is sufficient, but be sure you know who you're talking to ahead of time
• volunteer service (organizing confs and workshops, etc.) also make a good candidate and serves networking purposes
• other conferences to consider: computer vision and machine learning (like NIPS)
• teaching experience is important for being a good candidate. There are varying degrees of engagement here: TAing, running reading groups, the teaching certificate program Warren did

From Peter Richardson:

• anticipate future hot topics (around the time you expect to get your PhD). Resolution and scanning technology are both improving for diffusion MRI, and techniques that anticipate or are adaptable to this change have future value
• Peter recommends the "ambush" model of clinical publications: build up preliminary work as a series of CS publications and then publish the big groundbreaker in a clinical journal
• keep a finger in the wind: work closely with clinical collaborators (but not just as a code monkey), get practice writing for medical audiences (by submitting to clinical journals/confs), go to medical/clinical confs (to learn the culture, community, jargon, and state of the art). Get to know the field and jump on open problems or (undeservedly) stagnant sub-areas.
• research (as well as SW developed for research) should follow a spiral development model rather than a waterfall: get preliminary versions of everything early. Spin off publications as you iterate. Research and development simply can't be strictly sequential; you don't know enough about the future to plan it unless you've already experienced it some.
• make detours liberally. Don't pigeonhole yourself or turn down exciting ideas because they're "not your area". Again, keep a finger in the wind
• a diverse portfolio of research keeps you resistant to scooping, but so does being aware of the field, both its CS and its medical/clinical aspects
• paper style: be aware of the fact that English is not the first language for most of your audience

From David:

• don't release all code online; little throwaway things don't belong. But anything that solves a generic problem is useful and belongs in the public eye
• what is a faculty job? Teaching, research, service to your discipline, service to your department, service to organizations.
• be careful not to drive things by conference deadlines
• fast iterations; don't waterfall! Think hard about how you can do a little bit of each phase of a long-term research project all the time rather than doing each phase completely in order.

From David 2009-01-12:

• Conference and journal submissions should be "trimmed off" the main thrust of your fundamental research. They should pull your progress from the sides, not from the top.