A variational perspective on accelerated methods in optimization
From MaRDI portal
Abstract: Accelerated gradient methods play a central role in optimization, achieving optimal rates in many settings. While many generalizations and extensions of Nesterov's original acceleration method have been proposed, it is not yet clear what is the natural scope of the acceleration concept. In this paper, we study accelerated methods from a continuous-time perspective. We show that there is a Lagrangian functional that we call the emph{Bregman Lagrangian} which generates a large class of accelerated methods in continuous time, including (but not limited to) accelerated gradient descent, its non-Euclidean extension, and accelerated higher-order gradient methods. We show that the continuous-time limit of all of these methods correspond to traveling the same curve in spacetime at different speeds. From this perspective, Nesterov's technique and many of its generalizations can be viewed as a systematic way to go from the continuous-time curves generated by the Bregman Lagrangian to a family of discrete-time accelerated algorithms.
Recommendations
- Multiscale analysis of accelerated gradient methods
- A variational formulation of accelerated optimization on Riemannian manifolds
- Understanding the acceleration phenomenon via high-resolution differential equations
- A differential equation for modeling Nesterov's accelerated gradient method: theory and insights
- Generalized momentum-based methods: a Hamiltonian perspective
Cited in
(only showing first 100 items - show all)- Accelerations for a variety of global optimization methods
- Accelerations for global optimization covering methods using second derivatives
- Stochastic mirror descent dynamics and their convergence in monotone variational inequalities
- Accelerated schemes for a class of variational inequalities
- Accelerating optimization by tracing valley
- Preconditioned accelerated gradient descent methods for locally Lipschitz smooth objectives with applications to the solution of nonlinear PDEs
- Selection dynamics for deep neural networks
- Is there an analog of Nesterov acceleration for gradient-based MCMC?
- Accelerated information gradient flow
- Revisiting the ODE method for recursive algorithms: fast convergence using quasi stochastic approximation
- Iterative ensemble Kalman methods: a unified perspective with some new variants
- How does momentum benefit deep neural networks architecture design? A few case studies
- Understanding the acceleration phenomenon via high-resolution differential equations
- From differential equation solvers to accelerated first-order methods for convex optimization
- A control-theoretic perspective on optimal high-order optimization
- High-order symplectic Lie group methods on \(SO(n)\) using the polar decomposition
- Finding extremals of Lagrangian actions
- Fast primal-dual algorithm via dynamical system for a linearly constrained convex optimization problem
- Online optimization of switched LTI systems using continuous-time and hybrid accelerated gradient flows
- Triggered gradient tracking for asynchronous distributed optimization
- Generating Nesterov's accelerated gradient algorithm by using optimal control theory for optimization
- Accelerated optimization on Riemannian manifolds via discrete constrained variational integrators
- An \(O(s^r)\)-resolution ODE framework for understanding discrete-time algorithms and applications to the linear convergence of minimax problems
- Finding geodesics joining given points
- Discrete processes and their continuous limits
- Bregman Itoh-Abe methods for sparse optimisation
- User-friendly guarantees for the Langevin Monte Carlo with inaccurate gradient
- Exponential convergence of distributed primal-dual convex optimization algorithm without strong convexity
- Mass-spring-damper networks for distributed optimization in non-Euclidean spaces
- Continuous relaxations for the traveling salesman problem
- Fractional differential equation approach for convex optimization with convergence rate analysis
- Accelerated variational PDEs for efficient solution of regularized inversion problems
- PDE acceleration: a convergence rate analysis and applications to obstacle problems
- Exploring critical points of energy landscapes: from low-dimensional examples to phase field crystal PDEs
- Explicit stabilised gradient descent for faster strongly convex optimisation
- Proximal gradient flow and Douglas-Rachford splitting dynamics: global exponential stability via integral quadratic constraints
- Robust hybrid zero-order optimization algorithms with acceleration via averaging in time
- A differential equation for modeling Nesterov's accelerated gradient method: theory and insights
- Accelerating Proximal Markov Chain Monte Carlo by Using an Explicit Stabilized Method
- On dissipative symplectic integration with applications to gradient-based optimization
- Convergence rates of inertial primal-dual dynamical methods for separable convex optimization problems
- Search direction correction with normalized gradient makes first-order methods faster
- scientific article; zbMATH DE number 1268598 (Why is no real title available?)
- Stochastic Methods for Composite and Weakly Convex Optimization Problems
- On the convergence of gradient-like flows with noisy gradient input
- The approximate duality gap technique: a unified theory of first-order methods
- Stochastic modified equations and dynamics of stochastic gradient algorithms. I: Mathematical foundations
- Analysis of optimization algorithms via integral quadratic constraints: nonstrongly convex problems
- Projected dynamical systems on irregular, non-Euclidean domains for nonlinear optimization
- A general system of differential equations to model first-order adaptive algorithms
- Accelerated first-order primal-dual proximal methods for linearly constrained composite convex programming
- Accelerated extra-gradient descent: a novel accelerated first-order method
- The connections between Lyapunov functions for some optimization algorithms and differential equations
- Optimization with momentum: dynamical, control-theoretic, and symplectic perspectives
- A Lyapunov analysis of accelerated methods in optimization
- Unified acceleration of high-order algorithms under general Hölder continuity
- Implicit Regularization and Momentum Algorithms in Nonlinearly Parameterized Adaptive Control and Prediction
- Adaptive Hamiltonian variational integrators and applications to symplectic accelerated optimization
- An optimal high-order tensor method for convex optimization
- Essential convergence rate of ordinary differential equations appearing in optimization
- Accelerated Optimization in the PDE Framework: Formulations for the Manifold of Diffeomorphisms
- Potential Function-Based Framework for Minimizing Gradients in Convex and Min-Max Optimization
- Contractivity of Runge-Kutta methods for convex gradient systems
- Multiscale analysis of accelerated gradient methods
- A Continuous-Time Analysis of Distributed Stochastic Gradient
- Dynamical Systems Theory and Algorithms for NP-hard Problems
- Accelerated Optimization in the PDE Framework Formulations for the Active Contour Case
- Asymptotic analysis via stochastic differential equations of gradient descent algorithms in statistical and computational paradigms
- Hessian barrier algorithms for linearly constrained optimization problems
- Generalized momentum-based methods: a Hamiltonian perspective
- Conformal symplectic and relativistic optimization
- A variational formulation of accelerated optimization on Riemannian manifolds
- A primal-dual flow for affine constrained convex optimization
- Accelerated differential inclusion for convex optimization
- Inertial primal-dual dynamics with damping and scaling for linearly constrained convex optimization problems
- Neurodynamic optimization approaches with finite/fixed-time convergence for absolute value equations
- Last-iterate convergence of saddle-point optimizers via high-resolution differential equations
- An ordinary differential equation for modeling Halpern fixed-point Algorithm
- Novel projection neurodynamic approaches for constrained convex optimization
- Factor-\(\sqrt{2}\) acceleration of accelerated gradient methods
- Practical perspectives on symplectic accelerated optimization
- Deterministic neural networks optimization from a continuous and energy point of view
- Fast gradient method for low-rank matrix estimation
- A Systematic Approach to Lyapunov Analyses of Continuous-Time Models in Convex Optimization
- No-regret algorithms in on-line learning, games and convex optimization
- No-regret dynamics in the Fenchel game: a unified framework for algorithmic convex optimization
- Conformal mirror descent with logarithmic divergences
- A new dynamical system with self-adaptive dynamical stepsize for pseudomonotone mixed variational inequalities
- A Recursively Recurrent Neural Network (R2N2) Architecture for Learning Iterative Algorithms
- Time-adaptive Lagrangian variational integrators for accelerated optimization
- First-order methods for convex optimization
- Bregman dynamics, contact transformations and convex optimization
- From Halpern's fixed-point iterations to Nesterov's accelerated interpretations for root-finding problems
- A second order primal-dual dynamical system for a convex-concave bilinear saddle point problem
- Lagrangian and Hamiltonian dynamics for probabilities on the statistical bundle
- Accelerated Forward-Backward Optimization using Deep Learning
- Online accelerated data-driven learning for optimal feedback control of discrete-time partially uncertain systems
- Linear convergence of forward-backward accelerated algorithms without knowledge of the modulus of strong convexity
- The regularized submodular maximization via the Lyapunov method
- Differential equations in data analysis
This page was built for publication: A variational perspective on accelerated methods in optimization
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4646231)
