New algorithms reduce bilevel optimization complexity to ε^(-1.5).
problem Efficiently solving bilevel optimization problems in machine learning.
method Proposed two new algorithms: one using momentum-based recursive iterations, the other using recursive gradient estimations.
result Achieved computational complexity of ε^(-1.5), significantly faster than previous methods.
Learning to Optimize is a recently proposed framework for learning optimization algorithms using reinforcement learning. In this paper, we explore learning an optimization algorithm for training shallow neural nets. Such high-dimensional stochastic optimization problems present interesting challenges for existing reinf…
An optimal algorithm for multi-armed bandits with constraints.
problem Optimizing decisions in constrained multi-armed bandit problems.
method An index-based deterministic algorithm using Locatelli's anytime thresholding under known optimal value assumption.
result The algorithm achieves asymptotic optimality with probability approaching 1.
Paper proposes a reinforcement learning framework for efficient hyper-parameter tuning of stochastic optimization algorithms.
problem Efficient tuning of hyper-parameters for stochastic optimization algorithms.
method Modeling hyper-parameter tuning as a Markov decision process and using policy gradient algorithms.
result The proposed framework significantly reduces the time required for hyper-parameter tuning compared to Bayesian optimization.
New algorithms achieve optimal DP convex optimization with linear time and gradient computations.
problem Private stochastic convex optimization with optimal excess loss.
method Two new techniques: variable batch sizes and localization with stable optimization.
result Achieves optimal bound on excess loss with O ( min { n , n 2 / d } ) O(\min\{n, n^2/d\}) O ( min { n , n 2 / d }) gradient computations. Two new algorithms optimize decentralized convex optimization with reduced communication rounds.
problem Decentralized minimization of smooth strongly convex functions in a network.
method Proposes two new algorithms based on accelerated Forward Backward methods.
result First algorithm is optimal in terms of communication rounds and gradient computations.
MadEvolve optimizes trading algorithms using LLMs, achieving significant improvements in feature generation and trading strategy optimization.
problem Optimizing trading algorithms for better performance and feature generation.
method A framework inspired by Alpha-Evolve, using LLMs to evolve trading strategies and feature pipelines.
result Significant improvements in trading performance across various tasks, including feature generation and trading strategy optimization.
This thesis analyzes and improves convergence rates of bilevel optimization algorithms in machine learning.
problem Convergence analysis and algorithm design for bilevel optimization in machine learning.
method Comprehensive convergence rate analysis for both problem-based and algorithm-based bilevel optimization formulations.
result First lower bounds and matching upper bounds for bilevel optimization, and new stochastic algorithms with lower complexity.
Linear optimization is many times algorithmically simpler than non-linear convex optimization. Linear optimization over matroid polytopes, matching polytopes and path polytopes are example of problems for which we have simple and efficient combinatorial algorithms, but whose non-linear convex counterpart is harder and …
SMAC3 optimizes machine learning hyperparameters efficiently.
problem Optimizing hyperparameters for machine learning algorithms.
method Bayesian Optimization framework with facades for various use cases.
result Improves performance with minimal evaluations.
Study compares various optimization algorithms for deep learning.
problem Challenges in optimizing deep neural networks as they grow deeper and datasets increase.
method Implemented and compared adaptive gradient methods for supervised and unsupervised tasks on four image datasets.
result Differences in behavior and performance of optimization algorithms during training.
New algorithm tackles multimodal bandits with optimal regret bounds.
problem Stochastic multi-armed bandit problem with multimodal rewards.
method Proposed computationally tractable algorithm for multimodal rewards.
result Asymptotically optimal algorithms implemented for the bandit problem.
New algorithms ensure reproducibility and optimal convergence in convex optimization.
problem Trade-off between reproducibility and convergence rate in convex optimization.
method Regularization-based algorithms for smooth convex minimization and minimax optimization.
result Achieves optimal reproducibility and near-optimal gradient complexity for various oracle settings.
New algorithm reduces individual regret and communication costs in cooperative bandits.
problem Optimal individual and group regret in cooperative multi-agent bandits.
method Integrates a new communication policy into a learning algorithm.
result Achieves optimal individual regret and constant communication costs.
Many machine learning algorithms minimize a regularized risk, and stochastic optimization is widely used for this task. When working with massive data, it is desirable to perform stochastic optimization in parallel. Unfortunately, many existing stochastic optimization algorithms cannot be parallelized efficiently. In t…
Modified K-means ensures local optimality with same complexity.
problem Lack of rigorous analysis on local optimality guarantees of K-means.
method Proposed modifications to K-means ensuring local optimality.
result Proposed methods provide improved locally optimal solutions.
New algorithms achieve optimal robustness in stochastic convex optimization under contamination.
problem Determining optimal rates for robust stochastic convex optimization under ε ε ε -contamination. method Developed novel algorithms achieving minimax-optimal excess risk under ε ε ε -contamination model without stringent assumptions. result Achieved minimax-optimal excess risk (up to logarithmic factors) under ε ε ε -contamination model. Algorithm design is a laborious process and often requires many iterations of ideation and validation. In this paper, we explore automating algorithm design and present a method to learn an optimization algorithm, which we believe to be the first method that can automatically discover a better algorithm. We approach th…
Meta algorithm solves multivariate optimization using univariate optimizers.
problem Multivariate global optimization problems.
method Meta algorithm combining univariate global optimizers.
result Meta algorithm provides robust regret guarantees.
Adaptive algorithm AMSGrad converges for weakly convex constrained optimization problems.
problem Solving constrained stochastic optimization problems with weakly convex objectives.
method Analysis of AMSGrad algorithm for a specific class of problems.
result AMSGrad achieves a convergence rate of i l d e O ( t − 1 / 4 ) \mathcal{ ilde O}(t^{-1/4}) i l d e O ( t − 1/4 ) for the norm of the gradient of the Moreau envelope. PAC-Bayesian theory applied to learning optimization algorithms with generalization guarantees.
problem Learning optimization algorithms with provable generalization guarantees and explicit trade-offs.
method PAC-Bayes theory applied to learning-to-optimize, reformulating the learning procedure into a one-dimensional minimization problem.
result Learned optimization algorithms outperform deterministic worst-case analysis algorithms, even in the limit case of guaranteed convergence.
Simpler majority vote of three classifiers achieves optimal error bounds.
problem Developing an optimal PAC learning algorithm in the realizable setting.
method Returning the majority vote of three ERM classifiers.
result Achieves optimal in-expectation bound on error.
Optimizes pure exploration in linear bandits with a new algorithm.
problem Best-arm identification in linear stochastic bandits.
method Developed the first asymptotically optimal algorithm for fixed-confidence pure exploration in linear bandits.
result Avoids the pitfall of a simple but difficult instance and bypasses the need to solve an optimal design problem.
Paper tackles non-stationary kernelized bandits with near-optimal algorithm.
problem Minimizing regret in a time-varying reward function.
method Near-optimal algorithm with a novel restarting phased elimination with random permutation (R-PERP).
result Regret upper bound matches the lower bound, making the algorithm near-optimal.
New algorithm AG-OG optimizes separable convex-concave problems efficiently.
problem Efficiently solving separable convex-concave minimax optimization problems.
method Leverages Nesterov acceleration and optimistic gradient on component and coupling parts of the problem.
result Achieves optimal convergence rate for various settings including bilinearly coupled problems.
Two algorithms find optimal points in decentralized optimization.
problem Decentralized non-convex stochastic optimization with composite objective functions.
method Prox-DASA and Prox-DASA-GT algorithms for finding ε-stationary points.
result Achieves comparable complexity without large batch sizes or complex per-iteration operations.
Establishes geometric convergence of iterative optimization algorithms.
problem Analyzes convergence of iterative optimization algorithms under general assumptions.
method General framework for iterative optimization algorithms, proving asymptotic geometric convergence and providing convergence rates.
result Asymptotic geometric convergence of iterative optimization algorithms with exact rate.
Unified framework for analyzing online convex optimization across various settings.
problem Analyzing online convex optimization in different settings and feedback types.
method Unified framework allowing systematic proposal and analysis of meta-algorithms.
result Comparable regret bounds for various feedback types and adversary types.
We propose accelerated randomized coordinate descent algorithms for stochastic optimization and online learning. Our algorithms have significantly less per-iteration complexity than the known accelerated gradient algorithms. The proposed algorithms for online learning have better regret performance than the known rando…
We consider the closely related problems of bandit convex optimization with two-point feedback, and zero-order stochastic convex optimization with two function evaluations per round. We provide a simple algorithm and analysis which is optimal for convex Lipschitz functions. This improves on \cite{dujww13}, which only p…
The paper interprets policy-gradient algorithms using continuation theory.
problem Optimizing nonconvex functions in reinforcement learning.
method Formulates policy optimization as optimization by continuation, interprets policy-gradient algorithms as implicitly optimizing deterministic policies.
result Exploration in policy-gradient algorithms is seen as computing a continuation of the return of the policy.
Unified algorithm solves convex optimization problems with optimal rates.
problem Solving nonsmooth constrained convex optimization problems.
method Unified randomized block-coordinate primal-dual algorithm.
result Achieves optimal convergence rates of O ( n / k ) \mathcal{O}(n/k) O ( n / k ) and O ( n 2 / k 2 ) \mathcal{O}(n^2/k^2) O ( n 2 / k 2 ) . Paper develops momentum schemes with variance reduction for non-convex composition optimization.
problem Lack of convergence guarantee and efficient momentum design in existing algorithms.
method Develops various momentum schemes with SPIDER-based variance reduction.
result Achieves near-optimal sample complexity and linear convergence rate.
When a black-box optimization objective can only be evaluated with costly or noisy measurements, most standard optimization algorithms are unsuited to find the optimal solution. Specialized algorithms that deal with exactly this situation make use of surrogate models. These models are usually continuous and smooth, whi…
Study shows code-level optimizations significantly impact deep RL algorithms.
problem Understanding the impact of implementation details on deep RL algorithms.
method Case study on PPO and TRPO, investigating the effects of code-level optimizations.
result Code-level optimizations are crucial for performance in deep RL algorithms.
Bayesian optimization algorithm reduces regret with efficient region pruning.
problem Sequential optimization of unknown functions in high-dimensional spaces.
method Gaussian process-based, domain shrinking through tree-based region pruning.
result Order-optimal regret performance with reduced computational complexity.
New algorithm solves unbalanced optimal transport on trees in quasi-linear time.
problem Efficiently solving unbalanced optimal transport problems on trees.
method Proposed an algorithm that solves a more general unbalanced optimal transport problem exactly in quasi-linear time on a tree metric.
result Solves unbalanced optimal transport on trees in quasi-linear time (less than one second for a tree with one million nodes).
New algorithm achieves optimal regret in average reward MDPs without prior bias information.
problem Achieving optimal regret in average reward MDPs with computational efficiency and without prior bias information.
method Projective Mitigated Extended Value Iteration (PMEVI) to compute bias-constrained optimal policies efficiently.
result First tractable algorithm with minimax optimal regret of O ~ ( s p ( h ∗ ) S A T ) \widetilde{\mathrm{O}}(\sqrt{\mathrm{sp}(h^*) S A T}) O ( sp ( h ∗ ) S A T ) . Paper proposes a novel metric learning algorithm using Riemannian optimization.
problem Optimizing a smooth, convex function in Riemannian space with constraints.
method Developed a primal-dual algorithm with proximal operator for iterative optimization.
result Demonstrated the efficacy of the proposed metric learning algorithm on fund selection.
PPO algorithm converges to global optimality in multi-agent reinforcement learning.
problem Designing statistical guarantees for policy optimization methods in multi-agent reinforcement learning.
method Leveraging a multi-agent performance difference lemma, a localized action value function is used as a descent direction for each local policy, leading to a multi-agent PPO algorithm.
result The multi-agent PPO algorithm converges to the globally optimal policy at a sublinear rate under standard regularity conditions.
Optimizes algorithms for non-concave bandit problems.
problem Optimizing algorithms for non-concave bandit problems.
method Unified zeroth-order optimization paradigm.
result Minimax-optimal algorithms in the dimension for low-rank generalized linear bandit problems.
New algorithms achieve uniform stability for empirical risk minimization.
problem Designing uniformly stable optimization algorithms for empirical risk minimization.
method Black-box conversion of smooth optimization algorithms and development of Mirror Descent for smooth optimization.
result Optimal algorithms with uniform stability and convergence rates for smooth optimization.
New algorithm finds near-optimal policies efficiently in zero-sum games.
problem Lack of provable efficiency guarantees for policy optimization in zero-sum games.
method Policy optimization algorithm with function approximation.
result Proves efficient convergence to near-optimal policies with polynomial samples and iterations.
In this paper we study simulation based optimization algorithms for solving discrete time optimal stopping problems. This type of algorithms became popular among practioneers working in the area of quantitative finance. Using large deviation theory for the increments of empirical processes, we derive optimal convergenc…
New algorithms optimize without tuning, matching tuned SGD performance.
problem Optimizing machine learning models without manual hyperparameter tuning.
method Formalizes tuning-free algorithms for matching SGD performance with loose hints.
result Tuning-free algorithms can match SGD performance, but not optimal convergence rates.
New findings show many popular bandit algorithms are unstable, contradicting minimax optimality.
problem Challenges in statistical inference from bandit algorithms due to adaptive, non-i.i.d. nature.
method Analysis of stability properties of optimism-based bandit algorithms.
result Widely used minimax-optimal UCB-style algorithms are unstable.
Spectral algorithms solve optimal community detection and related problems.
problem Optimal detection of community structures and related substructures.
method Spectral algorithms applied to various planted substructures.
result Spectral algorithms achieve optimal performance for a wide range of planted substructures.
Parallel surrogate optimization algorithms have proven to be efficient methods for solving expensive noisy optimization problems. In this work we develop a new parallel surrogate optimization algorithm (ProSRS), using a novel tree-based "zoom strategy" to improve the efficiency of the algorithm. We prove that if ProSRS…