For maximum-likelihood IRL, the inner-problem Hessian at a realizable optimum equals the temperature-scaled trajectory Fisher matrix, which enables a scalable sketched hypergradient method.
Therefore, αDKL(epπθ ∥epϕ) =E τ∼epπθ " ∞X t=1 (αlogπ θ(at |s t)−r ϕ(st, at)) # +αlogZ ϕ
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Efficient Hypergradient Descent for Inverse Reinforcement Learning
For maximum-likelihood IRL, the inner-problem Hessian at a realizable optimum equals the temperature-scaled trajectory Fisher matrix, which enables a scalable sketched hypergradient method.