Amortized Implicit Differentiation for Stochastic Bilevel Optimization

Abstract

We study a class of algorithms for solving bilevel optimization problems in both stochastic and deterministic settings when the inner-level objective is strongly convex. Specifically, we consider algorithms based on inexact implicit differentiation and we exploit a warm-start strategy to amortize the estimation of the exact gradient. We then introduce a unified theoretical framework inspired by the study of singularly perturbed systems (Habets, 1974) to analyze such amortized algorithms. By using this framework, our analysis shows these algorithms to match the computational complexity of oracle methods that have access to an unbiased estimate of the gradient, thus outperforming many existing results for bilevel optimization. We illustrate these findings on synthetic experiments and demonstrate the efficiency of these algorithms on hyper-parameter optimization experiments involving several thousands of variables.

References in corpus (10)

• On Differentiating Parameterized Argmin and Argmax Problems with Application to Bi-level Optimization
• A Two-Timescale Framework for Bilevel Optimization: Complexity Analysis and Application to Actor-Critic
• Efficient and Modular Implicit Differentiation
• Bilevel Optimization: Convergence Analysis and Enhanced Design
• Finite Time Analysis of Linear Two-timescale Stochastic Approximation with Markovian Noise
• On the Iteration Complexity of Hypergradient Computation
• Estimate Sequences for Variance-Reduced Stochastic Composite Optimization
• Lower Bounds and Accelerated Algorithms for Bilevel Optimization
• Provably Faster Algorithms for Bilevel Optimization
• A Flexible Framework for Designing Trainable Priors with Adaptive Smoothing and Game Encoding