This article examines the implicit regularization effect of Stochastic Gradient Descent (SGD). We consider the case of SGD without replacement, the variant typically used to optimize large-scale neural networks. We analyze this algorithm in a more realistic regime than typically considered in theoretical works on SGD, as, e.g., we allow the product of the learning rate and Hessian to be $O(1)$. Our core theoretical result is that optimizing with SGD without replacement is locally equivalent to making an additional step on a novel regularizer. This implies that the trajectory of SGD without replacement diverges from both noise-injected GD and SGD with replacement (in which batches are sampled i.i.d.). Indeed, the two SGDs travel flat regions of the loss landscape in distinct directions and at different speeds. In expectation, SGD without replacement may escape saddles significantly faster and present a smaller variance. Moreover, we find that SGD implicitly regularizes the trace of the noise covariance in the eigendirections of small and negative Hessian eigenvalues. This coincides with penalizing a weighted trace of the Fisher Matrix and the Hessian on several vision tasks, thus encouraging sparsity in the spectrum of the Hessian of the loss in line with empirical observations from prior work. We also propose an explanation for why SGD does not train at the edge of stability (as opposed to GD).

We introduce a suite of new particle-based algorithms for sampling in constrained domains which are entirely learning rate free. Our approach leverages coin betting ideas from convex optimisation, and the viewpoint of constrained sampling as a mirrored optimisation problem on the space of probability measures. Based on this viewpoint, we also introduce a unifying framework for several existing constrained sampling algorithms, including mirrored Langevin dynamics and mirrored Stein variational gradient descent. We demonstrate the performance of our algorithms on a range of numerical examples, including sampling from targets on the simplex, sampling with fairness constraints, and constrained sampling problems in post-selection inference. Our results indicate that our algorithms achieve competitive performance with existing constrained sampling methods, without the need to tune any hyperparameters.

While ensuring stability for linear systems is well understood, it remains a major challenge for nonlinear systems. A general approach in such cases is to compute a combination of a Lyapunov function and an associated control policy. However, finding Lyapunov functions for general nonlinear systems is a challenging task. To address this challenge, several methods have been proposed that represent Lyapunov functions using neural networks. However, such approaches either focus on continuous-time systems, or highly restricted classes of nonlinear dynamics. We propose the first approach for learning neural Lyapunov control in a broad class of discrete-time systems. Three key ingredients enable us to effectively learn provably stable control policies. The first is a novel mixed-integer linear programming approach for verifying the discrete-time Lyapunov stability conditions, leveraging the particular structure of these conditions. The second is a novel approach for computing verified sublevel sets. The third is a heuristic gradient-based method for quickly finding counterexamples to significantly speed up Lyapunov function learning. Our experiments on four standard benchmarks demonstrate that our approach significantly outperforms state-of-the-art baselines. For example, on the path tracking benchmark, we outperform recent neural Lyapunov control baselines by an order of magnitude in both running time and the size of the region of attraction, and on two of the four benchmarks (cartpole and PVTOL), ours is the first automated approach to return a provably stable controller. Our code is available at: https://github.com/jlwu002/nlc_discrete.

The trained NLF can then be used for the stability Stability assessment of non-linear systems and ensuring their estimation of the real-world system. However, this approach safe and reliable operation are of paramount importance in will fail if the real-world system dynamics is different from any real-world engineering system. While learning-based the model used for training the NLF. At the same time, the control schemes have received a lot of attention recently, real-world system model can be different from the model estimated the lack of stability guarantees is a fundamental issue that from the collected data due to various reasons, such as prevents their wide-scale deployment in the real world. One estimation error and changes in the system parameters over standard approach to estimate the stability region of a general time. Repeating the training procedure every time whenever nonlinear system is to first find a Lyapunov function for the there is such a parametric mismatch turns impractical due system and characterize its region of attraction (ROA) as the to the unavailability of necessary data samples and the need stability region (Khalil 2015). A closed-loop system is stable to get a quick stability assessment. Thus, learning a neural in the sense of Lyapunov if the system trajectory converges Lyapunov function for a real-world system using only a small to the origin as long as the initial condition is inside the number of data samples and through a few gradient updates, ROA. The sum-of-squares approach is one popular method remains an open problem.

Lowering the memory requirement in full-parameter training on large models has become a hot research area. MeZO fine-tunes the large language models (LLMs) by just forward passes in a zeroth-order SGD optimizer (ZO-SGD), demonstrating excellent performance with the same GPU memory usage as inference. However, the simulated perturbation stochastic approximation for gradient estimate in MeZO leads to severe oscillations and incurs a substantial time overhead. Moreover, without momentum regularization, MeZO shows severe over-fitting problems. Lastly, the perturbation-irrelevant momentum on ZO-SGD does not improve the convergence rate. This study proposes ZO-AdaMU to resolve the above problems by adapting the simulated perturbation with momentum in its stochastic approximation. Unlike existing adaptive momentum methods, we relocate momentum on simulated perturbation in stochastic gradient approximation. Our convergence analysis and experiments prove this is a better way to improve convergence stability and rate in ZO-SGD. Extensive experiments demonstrate that ZO-AdaMU yields better generalization for LLMs fine-tuning across various NLP tasks than MeZO and its momentum variants.

We present Constrained Stein Variational Trajectory Optimization (CSVTO), an algorithm for performing trajectory optimization with constraints on a set of trajectories in parallel. We frame constrained trajectory optimization as a novel form of constrained functional minimization over trajectory distributions, which avoids treating the constraints as a penalty in the objective and allows us to generate diverse sets of constraint-satisfying trajectories. Our method uses Stein Variational Gradient Descent (SVGD) to find a set of particles that approximates a distribution over low-cost trajectories while obeying constraints. CSVTO is applicable to problems with arbitrary equality and inequality constraints and includes a novel particle resampling step to escape local minima. By explicitly generating diverse sets of trajectories, CSVTO is better able to avoid poor local minima and is more robust to initialization. We demonstrate that CSVTO outperforms baselines in challenging highly-constrained tasks, such as a 7DoF wrench manipulation task, where CSVTO succeeds in 20/20 trials vs 13/20 for the closest baseline. Our results demonstrate that generating diverse constraint-satisfying trajectories improves robustness to disturbances and initialization over baselines.

This paper studies the Partial Optimal Transport (POT) problem between two unbalanced measures with at most $n$ supports and its applications in various AI tasks such as color transfer or domain adaptation. There is hence the need for fast approximations of POT with increasingly large problem sizes in arising applications. We first theoretically and experimentally investigate the infeasibility of the state-of-the-art Sinkhorn algorithm for POT due to its incompatible rounding procedure, which consequently degrades its qualitative performance in real world applications like point-cloud registration. To this end, we propose a novel rounding algorithm for POT, and then provide a feasible Sinkhorn procedure with a revised computation complexity of $\mathcal{\widetilde O}(n^2/\varepsilon^4)$. Our rounding algorithm also permits the development of two first-order methods to approximate the POT problem. The first algorithm, Adaptive Primal-Dual Accelerated Gradient Descent (APDAGD), finds an $\varepsilon$-approximate solution to the POT problem in $\mathcal{\widetilde O}(n^{2.5}/\varepsilon)$, which is better in $\varepsilon$ than revised Sinkhorn. The second method, Dual Extrapolation, achieves the computation complexity of $\mathcal{\widetilde O}(n^2/\varepsilon)$, thereby being the best in the literature. We further demonstrate the flexibility of POT compared to standard OT as well as the practicality of our algorithms on real applications where two marginal distributions are unbalanced.

We study convergence rates of loss and uncertainty-based active learning algorithms under various assumptions. First, we provide a set of conditions under which a convergence rate guarantee holds, and use this for linear classifiers and linearly separable datasets to show convergence rate guarantees for loss-based sampling and different loss functions. Second, we provide a framework that allows us to derive convergence rate bounds for loss-based sampling by deploying known convergence rate bounds for stochastic gradient descent algorithms. Third, and last, we propose an active learning algorithm that combines sampling of points and stochastic Polyak's step size. We show a condition on the sampling that ensures a convergence rate guarantee for this algorithm for smooth convex loss functions. Our numerical results demonstrate efficiency of our proposed algorithm.

Black-box variational inference is widely used in situations where there is no proof that its stochastic optimization succeeds. We suggest this is due to a theoretical gap in existing stochastic optimization proofs: namely the challenge of gradient estimators with unusual noise bounds, and a composite non-smooth objective. For dense Gaussian variational families, we observe that existing gradient estimators based on reparameterization satisfy a quadratic noise bound and give novel convergence guarantees for proximal and projected stochastic gradient descent using this bound. This provides rigorous guarantees that methods similar to those used in practice converge on realistic inference problems.

This paper jointly considers privacy preservation and Byzantine-robustness in decentralized learning. In a decentralized network, honest-but-curious agents faithfully follow the prescribed algorithm, but expect to infer their neighbors' private data from messages received during the learning process, while dishonest-and-Byzantine agents disobey the prescribed algorithm, and deliberately disseminate wrong messages to their neighbors so as to bias the learning process. For this novel setting, we investigate a generic privacy-preserving and Byzantine-robust decentralized stochastic gradient descent (SGD) framework, in which Gaussian noise is injected to preserve privacy and robust aggregation rules are adopted to counteract Byzantine attacks. We analyze its learning error and privacy guarantee, discovering an essential tradeoff between privacy preservation and Byzantine-robustness in decentralized learning -- the learning error caused by defending against Byzantine attacks is exacerbated by the Gaussian noise added to preserve privacy. For a class of state-of-the-art robust aggregation rules, we give unified analysis of the "mixing abilities". Building upon this analysis, we reveal how the "mixing abilities" affect the tradeoff between privacy preservation and Byzantine-robustness. The theoretical results provide guidelines for achieving a favorable tradeoff with proper design of robust aggregation rules. Numerical experiments are conducted and corroborate our theoretical findings.