A new method for MARL with partial observations reduces communication overhead.
problem Inefficient MARL algorithms in large-scale problems due to state and action information sharing.
method Distributed zeroth-order policy optimization with local policy gradient estimation using consensus.
result The method converges to a policy that is a stationary point of the global objective function.
ZDPG learns model-free policies without critics, improving on PG.
problem Model-free policy learning in complex dynamic problems.
method Approximates policy-reward gradients via two-point stochastic evaluations of the Q-function.
result Restores true model-free policy learning without critics, with improved stability and efficiency.
ZOSPI improves RL policies with global value function exploitation.
problem Limited sample efficiency of PG algorithms in RL.
method Zeroth-order supervised policy improvement, leveraging global value function estimation.
result ZOSPI achieves competitive results with remarkable sample efficiency.
Paper develops methods to optimize policies directly from human feedback without reward inference.
problem Challenges in RLHF, including reward model overfitting and distribution shift.
method Develops two algorithms for RLHF without reward inference, using zeroth-order gradient approximators.
result Establishes polynomial convergence rates and outperforms existing methods in numerical experiments.
Zeroth-order optimization methods lack inherent privacy guarantees.
problem Ensuring differential privacy in zeroth-order optimization methods.
method Analyzing ZO-GD with and without random initialization for convex and strongly convex objectives.
result ZO-GD is not differentially private for strongly convex objectives and can have superlinear privacy loss.
New method for zeroth-order stochastic gradient algorithms provides confidence intervals.
problem Lack of inferential capabilities for zeroth-order stochastic gradient algorithms.
method Established central limit theorem and provided online estimators for asymptotic covariance matrix.
result Asymptotically valid confidence sets for parameter estimation and prediction.
ZSPO optimizes RL from unknown link functions using human feedback.
problem Designing RLHF algorithms for unknown link functions.
method Zero-order policy optimization with human preference feedback.
result ZSPO converges to a stationary policy with a polynomial rate.
In this paper, we propose and analyze zeroth-order stochastic approximation algorithms for nonconvex and convex optimization, with a focus on addressing constrained optimization, high-dimensional setting and saddle-point avoiding. To handle constrained optimization, we first propose generalizations of the conditional g…
This paper reviews zeroth-order optimization in signal processing and machine learning.
problem Optimization problems without gradient information.
method Iterative steps: gradient estimation, descent direction computation, solution update.
result Demonstrates applications in robustness evaluation and black-box model explanations.
Proximal gradient method has been playing an important role to solve many machine learning tasks, especially for the nonsmooth problems. However, in some machine learning problems such as the bandit model and the black-box learning problem, proximal gradient method could fail because the explicit gradients of these pro…
Improves zeroth-order optimization for private machine learning with public data.
problem High computation and memory cost of first-order DP methods.
method PAZO (Public Data Assisted Zeroth-order Optimization) framework.
result Achieves superior privacy/utility tradeoffs across tasks.
Bayesian optimization improves policy search in reinforcement learning.
problem Finding optimal policies with high variance estimates from random samples.
method Develops an algorithm combining Bayesian optimization and policy gradients.
result Improves sample complexity and reduces variance in empirical evaluations.
Two algorithms solve nonconvex minimax problems with linear constraints, achieving complexity guarantees.
problem Nonconvex minimax problems with coupled linear constraints.
method Zeroth-order primal-dual alternating projected gradient (ZO-PDAPG) and zeroth-order regularized momentum primal-dual projected gradient (ZO-RMPDPG) algorithms.
result Iteration complexity guarantees for solving nonconvex-(strongly) concave minimax problems with coupled linear constraints.
Discretizations of Langevin diffusions provide a powerful method for sampling and Bayesian inference. However, such discretizations require evaluation of the gradient of the potential function. In several real-world scenarios, obtaining gradient evaluations might either be computationally expensive, or simply impossibl…
New method improves zeroth-order stochastic optimization with adaptive sampling.
problem Optimization problems without gradient information.
method Adaptive sampling quasi-Newton method using finite differences.
result Significant improvement in performance with adaptive sample sizes.
We consider the problem of optimizing a high-dimensional convex function using stochastic zeroth-order queries. Under sparsity assumptions on the gradients or function values, we present two algorithms: a successive component/feature selection algorithm and a noisy mirror descent algorithm using Lasso gradient estimate…
Book covers tools for zeroth-order convex optimisation.
problem Zeroth-order convex optimisation.
method Cutting plane methods, interior point methods, continuous exponential weights, gradient descent, online Newton step.
result Improved existing bounds and algorithms.
Paper proposes ZO-NGD for more efficient black-box attacks.
problem Vulnerability of state-of-the-art DNNs to adversarial attacks.
method Zeroth-order natural gradient descent (ZO-NGD) for black-box attacks.
result ZO-NGD achieves significantly lower model query complexities.
Paper proposes algorithms for solving nonconvex-nonconcave problems with complexity guarantees.
problem Nonconvex-nonconcave minimax problems with PL condition.
method Zeroth-order AGDA and VRAGDA algorithms.
result Iteration complexities for obtaining ε-stationary points.
ConMeZO speeds up zeroth-order optimization for large language models.
problem Slow convergence in high-dimensional parameter spaces of large language models.
method Adaptive directional sampling in a cone centered around a momentum estimate.
result Achieves the same convergence rate as MeZO but up to 2X faster.
New algorithms solve complex minimax problems without needing derivatives.
problem Solving nonconvex-concave minimax problems efficiently.
method Zeroth-order alternating and proximal gradient algorithms.
result Iteration complexity and function value estimation bounds established.
We propose a method for zeroth order stochastic convex optimization that attains the suboptimality rate of O~(n7T−1/2) after T queries for a convex bounded function f:Rn→R. The method is based on a random walk (the \emph{Ball Walk}) on the epigraph of the function. Th…
Alternating direction method of multipliers (ADMM) is a popular optimization tool for the composite and constrained problems in machine learning. However, in many machine learning problems such as black-box attacks and bandit feedback, ADMM could fail because the explicit gradients of these problems are difficult or in…
Paper tackles gradient-free minimax optimization with variance reduction for faster convergence.
problem Gradient-free minimax optimization problems in machine learning.
method Variance reduction technique to design a novel zeroth-order gradient descent ascent algorithm.
result Achieves the best known query complexity of O(κ(d₁ + d₂)ε⁻³), outperforming previous methods.
New algorithm reduces dimensionality in stochastic optimization.
problem Stochastic optimization in high-dimensional problems.
method Proposes a sparsity-inducing stochastic gradient-free (SI-SGF) algorithm.
result Proves dimension-free query complexity in convex and strongly convex cases.
Zeroth-order methods favor flat minima in machine learning.
problem Finding solutions with small Hessian trace in optimization.
method Zeroth-order optimization with two-point estimator.
result Zeroth-order optimization converges to flat minima.
This paper analyzes and guarantees convergence of prior-guided ZO algorithms.
problem Understanding convergence properties of prior-guided zeroth-order optimization algorithms.
method Analysis of convergence under a greedy descent framework with various gradient estimators, and development of ARS algorithm.
result Convergence guarantee for prior-guided random gradient-free (PRGF) algorithms and accelerated random search (ARS) algorithm.
In this paper, we design and analyze a new zeroth-order online algorithm, namely, the zeroth-order online alternating direction method of multipliers (ZOO-ADMM), which enjoys dual advantages of being gradient-free operation and employing the ADMM to accommodate complex structured regularizers. Compared to the first-ord…
In this paper, we study zeroth-order algorithms for minimax optimization problems that are nonconvex in one variable and strongly-concave in the other variable. Such minimax optimization problems have attracted significant attention lately due to their applications in modern machine learning tasks. We first consider a …
LAZO reduces query complexity and variance in ZO methods.
problem High query complexity and variance in zeroth-order optimization.
method LAZO uses adaptive lazy queries to reduce variance and save queries.
result LAZO achieves lower regret and query complexity compared to existing methods.
New optimization method improves generalization across various tasks.
problem Improving zeroth-order optimization for better generalization.
method Exponential tilting objective to connect zeroth-order optimization with sharpness-aware minimization.
result Achieves better generalization compared to vanilla zeroth-order baselines.
Two types of zeroth-order stochastic algorithms have recently been designed for nonconvex optimization respectively based on the first-order techniques SVRG and SARAH/SPIDER. This paper addresses several important issues that are still open in these methods. First, all existing SVRG-type zeroth-order algorithms suffer …
A new method for optimizing functions without gradients, improving efficiency and convergence.
problem Optimizing functions without gradient information in machine learning.
method Hybrid Gradient Descent (HGE) using random and coordinate-wise gradient estimates.
result The proposed method achieves optimal convergence rates in convex cases and generalizes to non-convex cases.
Sparse perturbations improve convergence in SZO methods for faster training.
problem Dependency of SZO methods on function dimensionality limits their convergence speed.
method Sparse perturbations reduce the effective dimensionality of the optimization problem.
result Sparse SZO optimization leads to faster convergence in training loss and test accuracy.
New adaptive methods solve weakly convex stochastic optimization problems.
problem Solving weakly convex stochastic optimization problems.
method Adaptive first and zeroth-order methods using exponential moving averages.
result Established non-asymptotic convergence rates for nonsmooth and nonconvex problems.
As application demands for zeroth-order (gradient-free) optimization accelerate, the need for variance reduced and faster converging approaches is also intensifying. This paper addresses these challenges by presenting: a) a comprehensive theoretical analysis of variance reduced zeroth-order (ZO) optimization, b) a nove…
LORENZA improves LLM fine-tuning efficiency and generalization.
problem Improving robustness and generalization of LLMs under hardware constraints.
method AdaZo-SAM and LORENZA, combining Adam and SAM with zeroth-order estimation and randomized SVD.
result LORENZA achieves better generalization and reduced memory consumption compared to existing methods.
New PG losses improve decision optimization in misspecified models.
problem Improving decision optimization in models that are not perfectly specified.
method Introducing Perturbation Gradient (PG) losses to connect decision loss with directional derivatives and optimizing using gradient techniques.
result PG losses yield best-in-class policies asymptotically, even in misspecified settings.
DPZero fine-tunes large models privately without backpropagation.
problem Memory and privacy challenges in fine-tuning large language models.
method DPZero uses zeroth-order methods for private fine-tuning, avoiding backpropagation.
result DPZero achieves private fine-tuning of RoBERTa and OPT on various tasks.
Optimal algorithms for Riemannian optimization with reduced complexity.
problem Stochastic optimization on Riemannian manifolds with limited data.
method Zeroth-order Riemannian Averaging Stochastic Approximation algorithms using Riemannian moving-average estimators and novel geometric conditions.
result Achieves optimal sample complexities for generating approximate first-order stationary solutions.
MAZE attacks a model without data, using synthetic inputs.
problem Model Stealing attacks that require data to replicate a target model.
method MAZE uses zeroth-order gradient estimation to create synthetic inputs.
result MAZE achieves high clone accuracy without data.
Zeroth-order (a.k.a, derivative-free) methods are a class of effective optimization methods for solving complex machine learning problems, where gradients of the objective functions are not available or computationally prohibitive. Recently, although many zeroth-order methods have been developed, these approaches still…
A new hybrid-ordered SGD method reduces communication and complexity for non-convex optimization.
problem Balancing communication, computational complexity, and convergence rate in distributed non-convex optimization.
method Hybrid-ordered distributed SGD with pre-shared scalers and periodic vector communication.
result Order-wise faster convergence compared to existing methods.
In the learning to learn (L2L) framework, we cast the design of optimization algorithms as a machine learning problem and use deep neural networks to learn the update rules. In this paper, we extend the L2L framework to zeroth-order (ZO) optimization setting, where no explicit gradient information is available. Our lea…
New method estimates Riemannian derivatives from noisy function evaluations.
problem Optimizing functions on Riemannian manifolds with noisy data.
method Riemannian Gaussian smoothing for gradient and Hessian estimation.
result Oracle complexity independent of ambient dimension.
New analysis for black-box learning without gradients, improving generalization bounds.
problem Generalization error analysis for derivative-free optimization.
method Zeroth-order Stochastic Search (ZoSS) algorithm for Lipschitz and smooth losses.
result Generalization bounds independent of model dimension, batch size, and number of perturbed evaluations.
Stochastic zeroth-order (SZO), or gradient-free, optimization allows to optimize arbitrary functions by relying only on function evaluations under parameter perturbations, however, the iteration complexity of SZO methods suffers a factor proportional to the dimensionality of the perturbed function. We show that in scen…
Zeroth-order optimization is an important research topic in machine learning. In recent years, it has become a key tool in black-box adversarial attack to neural network based image classifiers. However, existing zeroth-order optimization algorithms rarely extract second-order information of the model function. In this…