Deep Learning Optimization

Overview

It studies how parameter updates, learning-rate and batch schedules, regularization, and model structure shape neural-network training. Common approaches range from SGD and AdamW to adaptive preconditioners such as Shampoo, curvature-aware and sharpness-aware methods, and matrix or spectral optimizers such as Muon; theory examines convergence, stability, and geometry, while systems work reduces memory, communication, and step time. Evaluation spans vision and tabular models, language-model pretraining, diffusion, and continual learning, measuring accuracy and loss alongside time to target and memory use. Optimizer rankings depend on architecture, batch size, training horizon, and compute budget, leaving a tension between per-step progress, wall-clock efficiency, generalization, and reliable tuning.

Latest papers

217 papers13 papers in the last four weeks, down 65% on the four weeks before. 0.1% of all new papers.

Sep 10, 2025

Predator-Prey Model: Driven Hunt for Accelerated Grokking

A machine learning method is proposed using two agents (representing two coupled replicas of the model parameters) that simulate the biological behavior of a predator and a prey. In this method, the two agents interact with each other via a pair of potentials that simulate the dynamics when the predator chases the prey while the prey runs away from the predator - to perform an optimization on the landscape. This method allows, for optimization landscapes with narrow ravines with gentle slope or even flat bottoms, to avoid optimization getting stuck in a ravine. For some examples of grokking (i.e., delayed generalization) problems we show that this method allows for achieving up to a hundred times faster learning in gradient evaluations and up to 60 times in the wall-clock time compared to the standard learning procedure.
Sep 3, 2025

Understanding and Improving Shampoo and SOAP via Kullback-Leibler Minimization

Shampoo and its efficient variant, SOAP, employ structured second-moment estimations and have shown strong performance for training neural networks (NNs). In practice, however, Shampoo typically requires step-size grafting with Adam to be competitive, and SOAP mitigates this by applying Adam in Shampoo's eigenbasis -- at the cost of additional memory overhead from Adam in both methods. Prior analyses have largely relied on the Frobenius norm to motivate these estimation schemes. We instead recast their estimation procedures as covariance estimation under Kullback-Leibler (KL) divergence minimization, revealing a previously overlooked theoretical limitation and motivating principled redesigns. Building on this perspective, we develop KL-Shampoo\textbf{KL-Shampoo} and KL-SOAP\textbf{KL-SOAP}, practical schemes that match or exceed the performance of Shampoo and SOAP in NN pre-training while achieving SOAP-level per-iteration runtime. Notably, KL-Shampoo does not rely on Adam to attain competitive performance, eliminating the memory overhead introduced by Adam. Across our experiments, KL-Shampoo consistently outperforms SOAP, Shampoo, and even KL-SOAP, establishing the KL-based approach as a promising foundation for designing structured methods in NN optimization. An implementation of KL-Shampoo/KL-SOAP is available at https://github.com/yorkerlin/KL-Methods
Sep 2, 2025

Fantastic Pretraining Optimizers and Where to Find Them

AdamW has long been the dominant optimizer in language model pretraining, despite numerous claims that alternative optimizers offer 1.4 to 2x speedup. We posit that two methodological shortcomings have obscured fair comparisons and hindered practical adoption: (i) unequal hyperparameter tuning and (ii) limited or misleading evaluation setups. To address these two issues, we conduct a systematic study of ten deep learning optimizers across four model scales (0.1B-1.2B parameters) and data-to-model ratios (1-8x the Chinchilla optimum). We find that fair and informative comparisons require rigorous hyperparameter tuning and evaluations across a range of model scales and data-to-model ratios, performed at the end of training. First, optimal hyperparameters for one optimizer may be suboptimal for another, making blind hyperparameter transfer unfair. Second, the actual speedup of many proposed optimizers over well-tuned baselines is lower than claimed and decreases with model size to only 1.1x for 1.2B parameter models. Thirdly, comparing intermediate checkpoints before reaching the target training budgets can be misleading, as rankings between two optimizers can flip during training due to learning rate decay. Through our thorough investigation, we find that all the fastest optimizers such as Muon and Soap, use matrices as preconditioners -- multiplying gradients with matrices rather than entry-wise scalars. However, the speedup of matrix-based optimizers is inversely proportional to model scale, decreasing from 1.4x over AdamW for 0.1B parameter models to merely 1.1x for 1.2B parameter models.
May 27, 2025

Position: Adopt Constraints Over Fixed Penalties in Deep Learning

Deep learning models are increasingly subject to explicit requirements, or constraints, such as safety levels, fairness limits, and sparsity budgets. In practice, such requirements are often handled through fixed weighted-sum penalization: the constraints are added to the task loss with fixed coefficients, and the resulting scalarized objective is minimized. This position paper argues that fixed penalization is often ill-suited to deep learning problems with non-negotiable requirements, for three main reasons. First, in non-convex settings, the penalized and constrained problems are not, in general, equivalent, so a minimizer of the penalized objective need not solve the constrained problem. Second, a penalty coefficient sets an objective-constraint trade-off rather than the constraint level, so finding an acceptable coefficient often takes trial and error, each trial a full training run. Third, fixed penalization weakens hard requirements into soft penalties to be traded off against task performance. As a result, compliance is no longer built into the problem, and requirements can be quietly relaxed. We therefore argue that, when a deep learning problem specifies non-negotiable requirements, one should solve the constrained problem itself instead of a surrogate defined by fixed penalization.
Mar 14, 2025

Asynchronous Sharpness-Aware Minimization For Fast and Accurate Deep Learning

Sharpness-Aware Minimization (SAM) is an optimization method that improves the generalization performance of machine learning models. Despite its effectiveness, it has not been widely adopted in real-world applications due to its high computational cost. Existing approaches reduce this overhead by approximating or reusing gradients in the perturbation step, but still assume sequential perturbation and update steps. In this study, we introduce AsyncSAM, a variant of SAM that enables concurrent model perturbation and parameter updates across multiple compute resources. By adjusting the perturbation batch size in a system-aware manner, our asynchronous scheme fully overlaps perturbation with updates, eliminating the overhead and allowing even CPUs to perform perturbations without extra time. We also provide theoretical analysis showing that AsyncSAM achieves the same order of convergence as conventional SAM. Extensive experiments demonstrate that AsyncSAM delivers superior generalization performance and the highest training throughput among state-of-the-art SAM variants. For example, in ViT fine-tuning on CIFAR-100, it achieves accuracy comparable to SAM while maintaining training time nearly identical to SGD.
Feb 15, 2025

Preconditioned Inexact Stochastic ADMM for Deep Model

Deep learning models are usually trained with stochastic gradient descent-based algorithms, but these optimizers face inherent limitations, such as slow convergence and stringent assumptions for convergence. In particular, data heterogeneity arising from distributed settings poses significant challenges to their theoretical and numerical performance. This paper develops an algorithm, PISA (Preconditioned Inexact Stochastic Alternating Direction Method of Multipliers). Grounded in rigorous theoretical guarantees, the algorithm converges under the sole assumption of Lipschitz continuity of the gradient on a bounded region, thereby removing the need for other conditions commonly imposed by stochastic methods. This capability enables the proposed algorithm to tackle the challenge of data heterogeneity effectively. Moreover, the algorithmic architecture enables scalable parallel computing and supports various preconditions, such as second-order information, second moment, and orthogonalized momentum by Newton-Schulz iterations. Incorporating the latter two preconditions in PISA yields two computationally efficient variants: SISA and NSISA. Comprehensive experimental evaluations for training or fine-tuning diverse deep models, including vision models, large language models, reinforcement learning models, generative adversarial networks, and recurrent neural networks, demonstrate superior numerical performance of SISA and NSISA compared to various state-of-the-art optimizers.
Nov 12, 2024

Convergence Rate Analysis of LION

The LION (evoLved sIgn mOmeNtum) optimizer for deep neural network training was found by Google via program search, with the simple sign update yet showing impressive performance in training large scale networks. Although previous studies have investigated its convergence properties, a comprehensive analysis, especially the convergence rate, is still desirable. Recognizing that LION can be regarded as solving a specific constrained problem, this paper focuses on demonstrating its convergence to the Karush-Kuhn-Tucker (KKT) point at the rate of O(dK−1/4)\cal O(\sqrt{d}K^{-1/4}) measured by gradient ℓ1\ell_1 norm, where dd is the problem dimension and KK is the number of iteration steps. Step further, we remove the constraint and establish that LION converges to the critical point of the general unconstrained problem at the same rate. This rate not only delivers the currently optimal dependence on the problem dimension dd but also tightly matches the theoretical lower bound for nonconvex stochastic optimization algorithms, which is typically measured using the gradient ℓ2\ell_2 norm, with respect to the number of iterations KK. Through extensive experiments, we not only demonstrate that LION achieves lower loss and higher performance compared to standard SGD, but also empirically confirm that the gradient ℓ1/ℓ2\ell_1/\ell_2 norm ratio aligns with Θ(d)Θ(\sqrt{d}), thus proving that our convergence rate matches the theoretical lower bound with respect to dd in the empirical sense.