Reading plan from advisor Ondřej Bajgar (email, 2026-07-04). Parent: thesis-hub.
The plan
Build on Bajgar et al.’s Bayesian IRL line of work. The thesis extends a main paper (foundation version file held locally; newer version reframed onto variational inference instead of MCMC).
Suggested reading order:
- bajgar-2024-valuewalk — start here. Their earlier Bayesian IRL paper; same framework but simpler, MCMC-based. Good on-ramp if you already know MCMC.
- main-vi-paper — the current main paper, built on VI instead of MCMC. Skim — much overlaps with ValueWalk and the foundation paper.
- foundation-paper — older/simpler version of the main paper; the base we build on. Link/title to be added.
Background to review as needed (don’t try to master up front):
- variational-inference — the newer paper’s engine. Tutorial: blei-2017-vi-review (or any shorter intro).
- mcmc — ValueWalk’s engine (Hamiltonian Monte Carlo).
- gaussian-processes — worth refreshing.
Notes
- Advisor’s guidance: read in order, skip freely on the newer paper (it repeats the older one), fill in background just-in-time.
- Open offer: Ondřej will send targeted resources when I’m stuck on something — flag blockers to him.
- Summer school link from the email intentionally omitted.