299 papers
The paper introduces LoFT, a novel parameter-efficient fine-tuning method that aligns optimizer dynamics (momentum and variance) with full fine-tuning within a low-rank subspace, thereby eliminating the need for hyperparameter tuning and achieving performance comparable to full fine-tuning without increasing inference costs.
This paper introduces StablePCA, a distributionally robust framework for extracting shared low-dimensional representations from multi-source data by maximizing worst-case explained variance, and addresses its inherent nonconvexity through a convex relaxation solved by an efficient Mirror-Prox algorithm with global convergence guarantees and a data-dependent certificate for solution tightness.
The paper proposes OTAD, a novel two-step defense framework that combines optimal transport theory with convex integration to train deep neural networks that achieve both high accuracy and certified local Lipschitz robustness against adversarial attacks.
This paper establishes finite-sample guarantees for a cost-driven representation learning method that learns a latent dynamic model by predicting multi-step costs, enabling the derivation of near-optimal controllers for finite-horizon Linear Quadratic Gaussian (LQG) control problems without explicitly modeling observations or actions.
This paper proposes Generative Adversarial Regression (GAR), a minimax framework that learns conditional risk scenarios by training generators to align their policy-induced risk with real data across a broad class of policies, thereby outperforming existing baselines in preserving downstream risk metrics like Value-at-Risk and Expected Shortfall.
This paper proposes ALS-IRLS, a novel outlier-robust algorithm that integrates an innovation-level adaptive thresholding mechanism with an IRLS-based Huber cost function to significantly improve the accuracy of noise covariance estimation and state filtering in the presence of measurement outliers compared to conventional methods.
This paper proposes a novel two-step transfer learning framework utilizing bi-level optimization to train a universal feature-extractor and a task-specific domain-adapter, enabling high-quality image reconstruction in data-scarce scenarios by effectively leveraging diverse cross-domain data.
This paper proposes FedCEF, a novel federated learning algorithm that combines a decoupled proximal update scheme with error feedback and control variates to achieve communication-efficient, sublinear convergence for non-convex composite optimization on heterogeneous data under extreme compression.
The paper introduces PolyFormer, a physics-informed machine learning framework that transforms complex physical constraints into efficient polytopic reformulations, achieving massive computational speedups and memory reductions while maintaining high solution quality for scalable optimization problems.
This paper establishes finite-sample guarantees for cost-driven state representation learning in infinite-horizon time-invariant Linear Quadratic Gaussian (LQG) control by analyzing two approaches—explicit latent modeling and implicit MuZero-like dynamics—while introducing a key technical proof of persistency of excitation for a novel stochastic process arising from quadratic regression.
This paper introduces MARIGOLD, a unified bi-level optimization framework that leverages zeroth-order methods to efficiently solve multi-task learning problems by dynamically balancing task gradients without requiring access to all task gradients, thereby overcoming the computational inefficiency of existing MGDA-type approaches.
This paper introduces Joint MDPs (JMDPs), a formalism that augments standard MDPs with a multi-action sample transition model to specify the joint distribution of counterfactual one-step outcomes, enabling the derivation of Bellman operators and convergent dynamic programming algorithms for environments with coupled dynamics.
This paper establishes a rigorous mathematical framework for leader-follower linear-quadratic stochastic graphon games with state-dependent diffusion, proving the existence and uniqueness of solutions and constructing a Stackelberg-Nash equilibrium for the continuum follower system.
This paper presents a finite element discretization for elliptic distributed optimal control problems with boundary value tracking, reformulating the problem as a state-based variational form to derive optimal error estimates and fast solvers, which are validated through numerical experiments.
This paper introduces a zero-shot transferable solution method for parametric optimal control problems that utilizes function encoder policies to learn reusable neural basis functions offline, enabling efficient online adaptation to varying objectives with minimal computational overhead and near-optimal performance.
This paper establishes a general fixed-point analysis for graph splitting methods, deriving an explicit formula for their limit points when applied to normal cones of closed linear subspaces to unify and extend existing convergence results.
This paper establishes a verification theorem linking coupled Hamilton-Jacobi-Bellman Master equations to Nash equilibria in two-player ergodic nonzero-sum McKean-Vlasov games and demonstrates explicit solutions for Linear-Quadratic-Gaussian settings by exploiting the polynomial structure of the value functions.
This study proposes a novel learn-then-optimize framework that leverages geographic data and SHAP-based interpretability to guide an integer programming model for the strategic deployment of public access defibrillators, thereby enhancing out-of-hospital cardiac arrest survival rates.
This paper investigates the role of semismooth derivatives in nonsmooth numerical analysis by exploring their interplay with semismoothness* properties for multifunctions, particularly in the context of solution maps to parametric inclusions, and establishes their coincidence with generalized Jacobians almost everywhere along with consequences for strict proto-differentiability.
This paper establishes a solid mathematical foundation for constrained optimization in Hilbert spaces by introducing the concept of essential Lagrange multipliers and a novel decomposition framework for KKT systems, which yields sharp results on the existence and uniqueness of multipliers, clarifies the distinctions between finite and infinite-dimensional theories, and characterizes the convergence of the augmented Lagrangian method.