REVIEW 2 cited by
Lower Bounds and Accelerated Algorithms for Bilevel Optimization
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Lower Bounds and Accelerated Algorithms for Bilevel Optimization
read the original abstract
Bilevel optimization has recently attracted growing interests due to its wide applications in modern machine learning problems. Although recent studies have characterized the convergence rate for several such popular algorithms, it is still unclear how much further these convergence rates can be improved. In this paper, we address this fundamental question from two perspectives. First, we provide the first-known lower complexity bounds of $\widetilde{\Omega}(\frac{1}{\sqrt{\mu_x}\mu_y})$ and $\widetilde \Omega\big(\frac{1}{\sqrt{\epsilon}}\min\{\frac{1}{\mu_y},\frac{1}{\sqrt{\epsilon^{3}}}\}\big)$ respectively for strongly-convex-strongly-convex and convex-strongly-convex bilevel optimizations. Second, we propose an accelerated bilevel optimizer named AccBiO, for which we provide the first-known complexity bounds without the gradient boundedness assumption (which was made in existing analyses) under the two aforementioned geometries. We also provide significantly tighter upper bounds than the existing complexity when the bounded gradient assumption does hold. We show that AccBiO achieves the optimal results (i.e., the upper and lower bounds match up to logarithmic factors) when the inner-level problem takes a quadratic form with a constant-level condition number. Interestingly, our lower bounds under both geometries are larger than the corresponding optimal complexities of minimax optimization, establishing that bilevel optimization is provably more challenging than minimax optimization.
Forward citations
Cited by 2 Pith papers
-
Pseudospectral Bounds for Transient Amplification in Coupled Gradient Descent
Pseudospectral bounds are proven for the Kreiss constant of block-triangular coupled Jacobians with symmetric diagonal blocks, yielding an O(K(J)^2 log(1/δ)) iteration complexity for stochastic coupled descent.
-
Pseudospectral Bounds for Transient Amplification in Coupled Gradient Descent
For block-triangular Jacobians with symmetric diagonal blocks of spectral radius at most γ, the Kreiss constant is at most 2 under weak coupling and at most 2+(||C||−2(1−γ))²/(4(1−γ)||C||) under strong coupling, contr...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.