paper-with-me

홈 › Papers

Jorge: Approximate Preconditioning for GPU-efficient Second-order Optimization

2023-10-18 · Siddharth Singh, Zachary Sating, Abhinav Bhatele

Despite their better convergence properties compared to first-order optimizers, second-order optimizers for deep learning have been less popular due to their significant computational costs. The primary efficiency bottleneck in such optimizers is matrix inverse calculations in the preconditioning step, which are expensive to compute on GPUs. In this paper, we introduce Jorge, a second-order optimizer that promises the best of both worlds -- rapid convergence benefits of second-order methods, and high computational efficiency typical of first-order methods. We address the primary computational bottleneck of computing matrix inverses by completely eliminating them using an approximation of the preconditioner computation. This makes Jorge extremely efficient on GPUs in terms of wall-clock time. Further, we describe an approach to determine Jorge's hyperparameters directly from a well-tuned SGD baseline, thereby significantly minimizing tuning efforts. Our empirical evaluations demonstrate the distinct advantages of using Jorge, outperforming state-of-the-art optimizers such as SGD, AdamW, and Shampoo across multiple deep learning models, both in terms of sample efficiency and wall-clock time.

📄 PDF Abstract BibTeX arXiv:2310.12298

Code (0)

등록된 구현이 없습니다.

Tasks

Computational EfficiencyGPUSecond-order methods

Methods 이 논문이 사용한 방법론

AdamW AdamW is a stochastic optimization method that modifies the typical implementation of weight decay in Adam, by decoupling [weight…
SGD Stochastic Gradient Descent is an iterative optimization technique that uses minibatches of data to form an expectation of the gradient, rather than the full gradient using…

Similar Papers 제목 키워드 기반

NysAct: A Scalable Preconditioned Gradient Descent using Nystrom Approximation

2025-06-10 · Hyunseok Seung, Jaewoo Lee, Hyunsuk Ko

Adaptive gradient methods are computationally efficient and converge quickly, but they often suffer from poor generalization. In contrast, second-order methods enhance convergence and generalization but typically incur h…

Second-order methods

Scalable K-FAC Training for Deep Neural Networks with Distributed Preconditioning

2022-06-30 · Lin Zhang, Shaohuai Shi, Wei Wang, Bo Li

The second-order optimization methods, notably the D-KFAC (Distributed Kronecker Factored Approximate Curvature) algorithms, have gained traction on accelerating deep neural network (DNN) training on GPU clusters. Howeve…

GPU

Amortized Proximal Optimization

2022-02-28 · Juhan Bae, Paul Vicol, Jeff Z. HaoChen, Roger Grosse

We propose a framework for online meta-optimization of parameters that govern optimization, called Amortized Proximal Optimization (APO). We first interpret various existing neural network optimizers as approximate stoch…

image-classificationImage ClassificationImage ReconstructionMeta-Learning+1

Adaptive Learning Rate via Covariance Matrix Based Preconditioning for Deep Neural Networks

2016-05-31 · Yasutoshi Ida, Yasuhiro Fujiwara, Sotetsu Iwamura

Adaptive learning rate algorithms such as RMSProp are widely used for training deep neural networks. RMSProp offers efficient training since it uses first order gradients to approximate Hessian-based preconditioning. How…

Stochastic Optimization

The Potential of Second-Order Optimization for LLMs: A Study with Full Gauss-Newton

2025-10-10 · Natalie Abreu, Nikhil Vyas, Sham Kakade, Depen Morwani arxiv

Recent efforts to accelerate LLM pretraining have focused on computationally-efficient approximations that exploit second-order structure. This raises a key question for large-scale training: how much performance is forf…