How to do Systems's Research - my perspective.

Here are my thoughts/experiences which might help current/future students.

TL;DR. Participating in implementation challenges and pairing up helps research.

  • This work evolved due to our participation in DIMACS Implementation Challenge.     - participating in such challenges helps us find the right problem to work on. //Important step for a research scholar.    - Organisors curate the more challenging and more relevant inputs.     - Irrespective of competition results, it is a win-win for research. If we win, we will get recognised. If we do not, we get the baselines (winners' solution code)     - In either case, if we continue to work on it further, we will have a paper out of it.    - Our earlier Steiner work (a result of another implementation Challenge -- PACE) had an excellent leaderboard that quickly evaluated any new idea/heuristics we implemented.     - I wish we/I have a personal dashboard for every research problem we/I work on that also logs (with some scores/metrics) for every new heuristic/code version we attempt.
  • Pairing up on research works (within and outside the group) always help.     - We had many brainstorming sessions (with/without advisors) with Somesh.    - We had come up with several heuristics/ideas from those sessions. At times debugging too.     - That can push (to perform more) and pulls-up (whenever we feel low) each other. //It happened many times to me. Thanks a lot, Somesh.    - If everything goes well, we will have one more paper (at least) out of this CVRP work.