Get our free extension to see links to code for papers anywhere online!Free extension: code links for papers anywhere!Free add-on: See code for papers anywhere!

Aaron Defazio, Ashok Cutkosky, Harsh Mehta, Konstantin Mishchenko

Learning rate schedules used in practice bear little resemblance to those recommended by theory. We close much of this theory/practice gap, and as a consequence are able to derive new problem-adaptive learning rate schedules. Our key technical contribution is a refined analysis of learning rate schedules for a wide class of optimization algorithms (including SGD). In contrast to most prior works that study the convergence of the average iterate, we study the last iterate, which is what most people use in practice. When considering only worst-case analysis, our theory predicts that the best choice is the linear decay schedule: a popular choice in practice that sets the stepsize proportionally to $1 - t/T$, where $t$ is the current iteration and $T$ is the total number of steps. To go beyond this worst-case analysis, we use the observed gradient norms to derive schedules refined for any particular task. These refined schedules exhibit learning rate warm-up and rapid learning rate annealing near the end of training. Ours is the first systematic approach to automatically yield both of these properties. We perform the most comprehensive evaluation of learning rate schedules to date, evaluating across 10 diverse deep learning problems, a series of LLMs, and a suite of logistic regression problems. We validate that overall, the linear-decay schedule matches or outperforms all commonly used default schedules including cosine annealing, and that our schedule refinement method gives further improvements.

Via

Konstantin Mishchenko, Aaron Defazio

We consider the problem of estimating the learning rate in adaptive methods, such as Adagrad and Adam. We describe two techniques, Prodigy and Resetting, to provably estimate the distance to the solution $D$, which is needed to set the learning rate optimally. Our techniques are modifications of the D-Adaptation method for learning-rate-free learning. Our methods improve upon the convergence rate of D-Adaptation by a factor of $O(\sqrt{\log(D/d_0)})$, where $d_0$ is the initial estimate of $D$. We test our methods on 12 common logistic-regression benchmark datasets, VGG11 and ResNet-50 training on CIFAR10, ViT training on Imagenet, LSTM training on IWSLT14, DLRM training on Criteo dataset, VarNet on Knee MRI dataset, as well as RoBERTa and GPT transformer training on BookWiki. Our experimental results show that our approaches consistently outperform D-Adaptation and reach test accuracy values close to that of hand-tuned Adam.

Via

Ashok Cutkosky, Aaron Defazio, Harsh Mehta

We introduce a technique for tuning the learning rate scale factor of any base optimization algorithm and schedule automatically, which we call \textsc{mechanic}. Our method provides a practical realization of recent theoretical reductions for accomplishing a similar goal in online convex optimization. We rigorously evaluate \textsc{mechanic} on a range of large scale deep learning tasks with varying batch sizes, schedules, and base optimization algorithms. These experiments demonstrate that depending on the problem, \textsc{mechanic} either comes very close to, matches or even improves upon manual tuning of learning rates.

Via

Fabian Schaipp, Ruben Ohana, Michael Eickenberg, Aaron Defazio, Robert M. Gower

We present new adaptive learning rates that can be used with any momentum method. To showcase our new learning rates we develop MoMo and MoMo-Adam, which are SGD with momentum (SGDM) and Adam together with our new adaptive learning rates. Our MoMo methods are motivated through model-based stochastic optimization, wherein we use momentum estimates of the batch losses and gradients sampled at each iteration to build a model of the loss function. Our model also makes use of any known lower bound of the loss function by using truncation. Indeed most losses are bounded below by zero. We then approximately minimize this model at each iteration to compute the next step. For losses with unknown lower bounds, we develop new on-the-fly estimates of the lower bound that we use in our model. Numerical experiments show that our MoMo methods improve over SGDM and Adam in terms of accuracy and robustness to hyperparameter tuning for training image classifiers on MNIST, CIFAR10, CIFAR100, Imagenet32, DLRM on the Criteo dataset, and a transformer model on the translation task IWSLT14.

Via

Aaron Defazio, Konstantin Mishchenko

The speed of gradient descent for convex Lipschitz functions is highly dependent on the choice of learning rate. Setting the learning rate to achieve the optimal convergence rate requires knowing the distance D from the initial point to the solution set. In this work, we describe a single-loop method, with no back-tracking or line searches, which does not require knowledge of $D$ yet asymptotically achieves the optimal rate of convergence for the complexity class of convex Lipschitz functions. Our approach is the first parameter-free method for this class without additional multiplicative log factors in the convergence rate. We present extensive experiments for SGD and Adam variants of our method, where the method automatically matches hand-tuned learning rates across more than a dozen diverse machine learning problems, including large-scale vision and language problems. Our method is practical, efficient and requires no additional function value or gradient evaluations each step. An open-source implementation is available (https://github.com/facebookresearch/dadaptation).

Via

Aaron Defazio, Baoyu Zhou, Lin Xiao

The classical AdaGrad method adapts the learning rate by dividing by the square root of a sum of squared gradients. Because this sum on the denominator is increasing, the method can only decrease step sizes over time, and requires a learning rate scaling hyper-parameter to be carefully tuned. To overcome this restriction, we introduce GradaGrad, a method in the same family that naturally grows or shrinks the learning rate based on a different accumulation in the denominator, one that can both increase and decrease. We show that it obeys a similar convergence rate as AdaGrad and demonstrate its non-monotone adaptation capability with experiments.

Via

Robert M. Gower, Aaron Defazio, Michael Rabbat

We propose a new stochastic gradient method that uses recorded past loss values to reduce the variance. Our method can be interpreted as a new stochastic variant of the Polyak Stepsize that converges globally without assuming interpolation. Our method introduces auxiliary variables, one for each data point, that track the loss value for each data point. We provide a global convergence theory for our method by showing that it can be interpreted as a special variant of online SGD. The new method only stores a single scalar per data point, opening up new applications for variance reduction where memory is the bottleneck.

Via

Aaron Defazio, Samy Jelassi

We introduce MADGRAD, a novel optimization method in the family of AdaGrad adaptive gradient methods. MADGRAD shows excellent performance on deep learning optimization problems from multiple fields, including classification and image-to-image tasks in vision, and recurrent and bidirectionally-masked models in natural language processing. For each of these tasks, MADGRAD matches or outperforms both SGD and ADAM in test set performance, even on problems for which adaptive methods normally perform poorly.

Via

Samy Jelassi, Aaron Defazio

First-order stochastic optimization methods are currently the most widely used class of methods for training deep neural networks. However, the choice of the optimizer has become an ad-hoc rule that can significantly affect the performance. For instance, SGD with momentum (SGD+M) is typically used in computer vision (CV) and Adam is used for training transformer models for Natural Language Processing (NLP). Using the wrong method can lead to significant performance degradation. Inspired by the dual averaging algorithm, we propose Modernized Dual Averaging (MDA), an optimizer that is able to perform as well as SGD+M in CV and as Adam in NLP. Our method is not adaptive and is significantly simpler than Adam. We show that MDA induces a decaying uncentered $L_2$-regularization compared to vanilla SGD+M and hypothesize that this may explain why it works on NLP problems where SGD+M fails.

Via

Aaron Defazio

Momentum methods are now used pervasively within the machine learning community for training non-convex models such as deep neural networks. Empirically, they out perform traditional stochastic gradient descent (SGD) approaches. In this work we develop a Lyapunov analysis of SGD with momentum (SGD+M), by utilizing a equivalent rewriting of the method known as the stochastic primal averaging (SPA) form. This analysis is much tighter than previous theory in the non-convex case, and due to this we are able to give precise insights into when SGD+M may out-perform SGD, and what hyper-parameter schedules will work and why.

Via