Back to search
2111.07692

FAST ADJOINT DIFFERENTIATION OF CHAOS

Angxiu Ni

correcthigh confidence
Category
Not specified
Journal tier
Strong Field
Processed
Sep 28, 2025, 12:56 AM

Audit review

The paper proves (i) the adjoint shadowing lemma: ν is the unique bounded solution of the inhomogeneous adjoint recursion ν_{n-1}=f^*ν_n+dΦ_{n-1}, and uses this to express the shadowing contribution S.C.=ρ(νX) (Theorem 3) ; (ii) the fast adjoint formula for the unstable divergence div^u_σ X^u = L(ω)X + ε̃∇_{ẽ}X with ω=(1/J) ε̃_1∇_{ẽ}f^* (Theorem 4) ; and (iii) the decomposition δρ(Φ)=S.C.−U.C. with U.C.=lim_W ρ(ψ_W div^u_σ X_u) (Section 1.2 and 2.3) . The model’s solution derives the same three statements. For (i) it gives a variation-of-constants construction and uniqueness via exponential dichotomy—substantially the same mathematical content as the paper’s expansion/uniqueness argument . For (ii) it reaches the same fast adjoint divergence identity but motivates it via disintegration of the SRB measure on unstable leaves and the leafwise equation for θ=d_u log h; this differs in method from the paper, which proceeds through second-order tangent expansions of unstable cubes and then characterizes an adjoint operator to obtain Theorem 4 . For (iii) both use Ruelle’s response theory and leafwise integration by parts to isolate the unstable piece in the form used by the paper . Minor caveat: the model informally differentiates leafwise densities (h) and asserts d_u log J = ω; while consistent with the final identity, this step needs careful justification in general. Overall, conclusions match and proofs are compatible; the paper emphasizes an algorithmic, expansion-based route, while the model gives a more measure/disintegration-flavored sketch.

Referee report (LaTeX)

\textbf{Recommendation:} minor revisions

\textbf{Journal Tier:} strong field

\textbf{Justification:}

The manuscript develops an adjoint shadowing lemma and a fast adjoint divergence formula that together enable an efficient adjoint response algorithm under uniform hyperbolicity. The theoretical steps are consistent with standard SRB/response theory and are well-aligned with the computational goals. The main identities are derived cleanly, and the algorithmic implications are convincingly motivated. Minor clarifications on assumptions and certain definitions would benefit readers and strengthen rigor, but no major obstacles to publication are apparent.