AI Paper Atlas

Research map / Optimization / Optimizers

The Curse of Unrolling: Rate of Differentiating Through Optimization

Damien Scieur, Gauthier Gidel, Quentin Bertrand, Fabian Pedregosa

NeurIPS 2022 · Optimization › Optimizers · approach: Differentiating through inner loops · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use Sobolev-optimized residual polynomials instead of gradient descent or Chebyshev unrollingUnrolled differentiation · Breast Cancer, bodyfat, and synthetic ridge regressionJacobian suboptimalityImproved
Use gradient descent with h=2/(L+ℓ) rather than h=1/LGradient descent unrolling · Quadratic objectives with bounded spectrumasymptotic Jacobian convergence rate: ((1−κ)/(1+κ))^(t−1) vs. (1−κ)^(t−1)Improved
Use gradient descent with h=2/(L+ℓ) rather than h=1/LGradient descent unrolling · Quadratic objectives with bounded spectrumburn-in and peak Jacobian suboptimality: Oκ→0(1/κ||∂x₀(θ)−∂x⋆(θ)||F)Hurt
Use Chebyshev unrolling rather than gradient descentUnrolled differentiation · Quadratic objectives with bounded spectrumburn-in duration: t≈2√(1/κ) vs. t≈1/κMixed
Apply the two-phase analysis to logistic regression rather than a quadratic objectiveUnrolled differentiation · Synthetic and breast-cancer logistic regressionpresence of burn-in followed by convergenceNo clear effect
Apply the two-phase analysis without the commutativity assumptionUnrolled differentiation · Generalized ridge regression with diagonal matrix Dpresence of burn-in followed by convergenceNo clear effect
Choose a large gradient-descent step size for faster asymptotic convergenceGradient descent unrolling · Quadratic objectives with bounded spectrumearly Jacobian suboptimality: Oκ→0(1/κ||∂x₀(θ)−∂x⋆(θ)||F)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy