Publications by Volkan Cevher | EPFL Graph Search

UNDERGRAD: A Universal Black-Box Optimization Method with Almost Dimension-Free Convergence Rate Guarantees

Universal methods for optimization are designed to achieve theoretically optimal convergence rates without any prior knowledge of the problem’s regularity parameters or the accurarcy of the gradient oracle employed by the optimizer. In this regard, existin ...

2022

Convergence of adaptive algorithms for constrained weakly convex optimization

Volkan Cevher, Ahmet Alacaoglu, Yurii Malitskyi

We analyze the adaptive first order algorithm AMSGrad, for solving a constrained stochastic optimization problem with a weakly convex objective. We prove the

\tilde O(t^{-1/2})

rate of convergence for the squared norm of the gradient of Moreau envelope, ...

2021

Subquadratic Overparameterization for Shallow Neural Networks

Volkan Cevher, Thomas Michaelsen Pethick, Armin Eftekhari, Chaehwan Song

Overparameterization refers to the important phenomenon where the width of a neural network is chosen such that learning algorithms can provably attain zero loss in nonconvex training. The existing theory establishes such global convergence using various i ...

2021

A Plug-and-Play Deep Image Prior

Volkan Cevher, Fabian Ricardo Latorre Gomez, Thomas Sanchez, Zhaodong Sun

Deep image priors (DIP) offer a novel approach for the regularization that leverages the inductive bias of a deep convolutional architecture in inverse problems. However, the quality of DIP approaches often degrades when the number of iterations exceeds a ...

IEEE2021

Sifting through the Noise: Universal First-Order Methods for Stochastic Variational Inequalities

Volkan Cevher, Kimon Antonakopoulos, Thomas Michaelsen Pethick, Ali Kavis

We examine a flexible algorithmic framework for solving monotone variational inequalities in the presence of randomness and uncertainty. The proposed template encompasses a wide range of popular first-order methods, including dual averaging, dual extrapola ...

2021

Regret Minimization in Stochastic Non-Convex Learning via a Proximal-Gradient Approach

Volkan Cevher, Shaul Nadav Hallak

This paper develops a methodology for regret minimization with stochastic first-order oracle feedback in online, constrained, non-smooth, non-convex problems. In this setting, the minimization of external regret is beyond reach for first-order methods, and ...

JMLR-JOURNAL MACHINE LEARNING RESEARCH2021

Robust Inverse Reinforcement Learning under Transition Dynamics Mismatch

Volkan Cevher, Luca Viano, Yu-Ting Huang, Kamalaruban Parameswaran

We study the inverse reinforcement learning (IRL) problem under a transition dynamics mismatch between the expert and the learner. Specifically, we consider the Maximum Causal Entropy (MCE) IRL learner model and provide a tight upper bound on the learner’s ...

2021

Forward-reflected-backward method with variance reduction

Volkan Cevher, Ahmet Alacaoglu

We propose a variance reduced algorithm for solving monotone variational inequalities. Without assuming strong monotonicity, cocoercivity, or boundedness of the domain, we prove almost sure convergence of the iterates generated by the algorithm to a soluti ...

2021

A first-order primal-dual method with adaptivity to local smoothness

Volkan Cevher, Maria-Luiza Vladarean

We consider the problem of finding a saddle point for the convex-concave objective

\min_x \max_y f(x) + \langle Ax, y\rangle - g^*(y)

, where

f

is a convex function with locally Lipschitz gradient and

g

is convex and possibly non-smooth. We propose an ...

2021

STORM+: Fully Adaptive SGD with Momentum for Nonconvex Optimization

Volkan Cevher, Ali Kavis

In this work we investigate stochastic non-convex optimization problems wherethe objective is an expectation over smooth loss functions, and the goal is to find an approximate stationary point. The most popular approach to handling such problems is varianc ...

2021

Volkan Cevher

Graph Chatbot

Chat with Graph Search

UNDERGRAD: A Universal Black-Box Optimization Method with Almost Dimension-Free Convergence Rate Guarantees

Convergence of adaptive algorithms for constrained weakly convex optimization

Subquadratic Overparameterization for Shallow Neural Networks

A Plug-and-Play Deep Image Prior

Sifting through the Noise: Universal First-Order Methods for Stochastic Variational Inequalities

Regret Minimization in Stochastic Non-Convex Learning via a Proximal-Gradient Approach

Robust Inverse Reinforcement Learning under Transition Dynamics Mismatch

Forward-reflected-backward method with variance reduction

A first-order primal-dual method with adaptivity to local smoothness

STORM+: Fully Adaptive SGD with Momentum for Nonconvex Optimization

Regret Minimization in Stochastic Non-Convex Learning via a Proximal-Gradient Approach

UNDERGRAD: A Universal Black-Box Optimization Method with Almost Dimension-Free Convergence Rate Guarantees

A Plug-and-Play Deep Image Prior

Convergence of adaptive algorithms for constrained weakly convex optimization

Subquadratic Overparameterization for Shallow Neural Networks

Sifting through the Noise: Universal First-Order Methods for Stochastic Variational Inequalities

Robust Inverse Reinforcement Learning under Transition Dynamics Mismatch

Forward-reflected-backward method with variance reduction

A first-order primal-dual method with adaptivity to local smoothness

STORM+: Fully Adaptive SGD with Momentum for Nonconvex Optimization