Primal Wasserstein Imitation Learning: Transport, Rewards, and Their Limits
Build PWIL from probability couplings, derive greedy transport rewards, and distinguish exact Wasserstein optimization from a practical learning surrogate.
More notes12 articles
Characteristic Sets: When RDF Cardinality Is Exact, and When It Is Estimated
Derive star-query counts from predicate co-occurrence, then expose the multiplicity, correlation, and compression assumptions behind the estimator.
Handling owl:sameAs via Rewriting: Equality Without Fact Explosion
How quotient representations compress equality closure, why rules must also be rewritten, and how SPARQL multiplicities survive expansion.
Competitive Programming with Rust: Invariants, Complexity, and a Reliable Workflow
A practical path from input parsing and integer bounds to graph algorithms, ownership, and reproducible contest tooling.