Adaptive gradient optimizers like Adam(W) are the default training algorithms for many deep learning architectures, such as transformers. Their diagonal preconditioner is based on the gradient outer product which is incorporated into the parameter update via a square root. While these methods are often motivated as approximate second-order methods, the square root represents a fundamental difference. In this work, we investigate how the behavior of adaptive methods changes when we remove the root, i.e. strengthen their second-order motivation. Surprisingly, we find that such square-root-free adaptive methods close the generalization gap to SGD on convolutional architectures, while maintaining their root-based counterpart's performance on transformers. The second-order perspective also has practical benefits for the development of adaptive methods with non-diagonal preconditioner. In contrast to root-based counterparts like Shampoo, they do not require numerically unstable matrix square roots and therefore work well in low precision, which we demonstrate empirically. This raises important questions regarding the currently overlooked role of adaptivity for the success of adaptive methods.

Decentralized Federated Learning (DFL) has received significant recent research attention, capturing settings where both model updates and model aggregations -- the two key FL processes -- are conducted by the clients. In this work, we propose Decentralized Sporadic Federated Learning ($\texttt{DSpodFL}$), a DFL methodology which generalizes the notion of sporadicity in both of these processes, modeling the impact of different forms of heterogeneity that manifest in realistic DFL settings. $\texttt{DSpodFL}$ unifies many of the prominent decentralized optimization methods, e.g., distributed gradient descent (DGD), randomized gossip (RG), and decentralized federated averaging (DFedAvg), under a single modeling framework. We analytically characterize the convergence behavior of $\texttt{DSpodFL}$, showing, among other insights, that we can match a geometric convergence rate to a finite optimality gap under more general assumptions than in existing works. Through experiments, we demonstrate that $\texttt{DSpodFL}$ achieves significantly improved training speeds and robustness to variations in system parameters compared to the state-of-the-art.

We investigate the training dynamics of two-layer neural networks when learning multi-index target functions. We focus on multi-pass gradient descent (GD) that reuses the batches multiple times and show that it significantly changes the conclusion about which functions are learnable compared to single-pass gradient descent. In particular, multi-pass GD with finite stepsize is found to overcome the limitations of gradient flow and single-pass GD given by the information exponent (Ben Arous et al., 2021) and leap exponent (Abbe et al., 2023) of the target function. We show that upon re-using batches, the network achieves in just two time steps an overlap with the target subspace even for functions not satisfying the staircase property (Abbe et al., 2021). We characterize the (broad) class of functions efficiently learned in finite time. The proof of our results is based on the analysis of the Dynamical Mean-Field Theory (DMFT). We further provide a closed-form description of the dynamical process of the low-dimensional projections of the weights, and numerical experiments illustrating the theory.

We study the problem of parameter-free stochastic optimization, inquiring whether, and under what conditions, do fully parameter-free methods exist: these are methods that achieve convergence rates competitive with optimally tuned methods, without requiring significant knowledge of the true problem parameters. Existing parameter-free methods can only be considered ``partially'' parameter-free, as they require some non-trivial knowledge of the true problem parameters, such as a bound on the stochastic gradient norms, a bound on the distance to a minimizer, etc. In the non-convex setting, we demonstrate that a simple hyperparameter search technique results in a fully parameter-free method that outperforms more sophisticated state-of-the-art algorithms. We also provide a similar result in the convex setting with access to noisy function values under mild noise assumptions. Finally, assuming only access to stochastic gradients, we establish a lower bound that renders fully parameter-free stochastic convex optimization infeasible, and provide a method which is (partially) parameter-free up to the limit indicated by our lower bound.

Stochastic Gradient Descent (SGD) with adaptive steps is now widely used for training deep neural networks. Most theoretical results assume access to unbiased gradient estimators, which is not the case in several recent deep learning and reinforcement learning applications that use Monte Carlo methods. This paper provides a comprehensive non-asymptotic analysis of SGD with biased gradients and adaptive steps for convex and non-convex smooth functions. Our study incorporates time-dependent bias and emphasizes the importance of controlling the bias and Mean Squared Error (MSE) of the gradient estimator. In particular, we establish that Adagrad and RMSProp with biased gradients converge to critical points for smooth non-convex functions at a rate similar to existing results in the literature for the unbiased case. Finally, we provide experimental results using Variational Autoenconders (VAE) that illustrate our convergence results and show how the effect of bias can be reduced by appropriate hyperparameter tuning.

Stochastic approximation is a class of algorithms that update a vector iteratively, incrementally, and stochastically, including, e.g., stochastic gradient descent and temporal difference learning. One fundamental challenge in analyzing a stochastic approximation algorithm is to establish its stability, i.e., to show that the stochastic vector iterates are bounded almost surely. In this paper, we extend the celebrated Borkar-Meyn theorem for stability from the Martingale difference noise setting to the Markovian noise setting, which greatly improves its applicability in reinforcement learning, especially in those off-policy reinforcement learning algorithms with linear function approximation and eligibility traces. Central to our analysis is the diminishing asymptotic rate of change of a few functions, which is implied by both a form of strong law of large numbers and a commonly used V4 Lyapunov drift condition and trivially holds if the Markov chain is finite and irreducible.

We consider the problem of designing sample efficient learning algorithms for infinite horizon discounted reward Markov Decision Process. Specifically, we propose the Accelerated Natural Policy Gradient (ANPG) algorithm that utilizes an accelerated stochastic gradient descent process to obtain the natural policy gradient. ANPG achieves $\mathcal{O}({\epsilon^{-2}})$ sample complexity and $\mathcal{O}(\epsilon^{-1})$ iteration complexity with general parameterization where $\epsilon$ defines the optimality error. This improves the state-of-the-art sample complexity by a $\log(\frac{1}{\epsilon})$ factor. ANPG is a first-order algorithm and unlike some existing literature, does not require the unverifiable assumption that the variance of importance sampling (IS) weights is upper bounded. In the class of Hessian-free and IS-free algorithms, ANPG beats the best-known sample complexity by a factor of $\mathcal{O}(\epsilon^{-\frac{1}{2}})$ and simultaneously matches their state-of-the-art iteration complexity.

In-Context Learning (ICL) empowers Large Language Models (LLMs) with the capacity to learn in context, achieving downstream generalization without gradient updates but with a few in-context examples. Despite the encouraging empirical success, the underlying mechanism of ICL remains unclear, and existing research offers various viewpoints of understanding. These studies propose intuition-driven and ad-hoc technical solutions for interpreting ICL, illustrating an ambiguous road map. In this paper, we leverage a data generation perspective to reinterpret recent efforts and demonstrate the potential broader usage of popular technical solutions, approaching a systematic angle. For a conceptual definition, we rigorously adopt the terms of skill learning and skill recognition. The difference between them is skill learning can learn new data generation functions from in-context data. We also provide a comprehensive study on the merits and weaknesses of different solutions, and highlight the uniformity among them given the perspective of data generation, establishing a technical foundation for future research to incorporate the strengths of different lines of research.

Although convergence analyses of SGD with momentum for nonconvex functions have been provided While stochastic gradient descent (SGD) with momentum [11, 16, 39], none of them explain why convergence has fast convergence and excellent generalizability, a is faster than with SGD. The generalizability of SGD with theoretical explanation for this is lacking. In this paper, momentum has been well studied, and various experimental we show that SGD with momentum smooths the objective findings have been reported. While it has been suggested function, the degree of which is determined by the learning that momentum plays a role in reducing stochastic noise rate, the batch size, the momentum factor, the variance of [8, 11], stochastic noise has been shown to increase generalizability the stochastic gradient, and the upper bound of the gradient [18, 37, 59], and it has been claimed that stochastic norm. This theoretical finding reveals why momentum improves noise can help an algorithm escape from local solutions generalizability and provides new insights into the with poor generalizability [10, 15, 20, 28, 32].

We give quantitative estimates for the rate of convergence of Riemannian stochastic gradient descent (RSGD) to Riemannian gradient flow and to a diffusion process, the so-called Riemannian stochastic modified flow (RSMF). Using tools from stochastic differential geometry we show that, in the small learning rate regime, RSGD can be approximated by the solution to the RSMF driven by an infinite-dimensional Wiener process. The RSMF accounts for the random fluctuations of RSGD and, thereby, increases the order of approximation compared to the deterministic Riemannian gradient flow. The RSGD is build using the concept of a retraction map, that is, a cost efficient approximation of the exponential map, and we prove quantitative bounds for the weak error of the diffusion approximation under assumptions on the retraction map, the geometry of the manifold, and the random estimators of the gradient.