A Second look at Exponential and Cosine Step Sizes: Simplicity, Adaptivity, and Performance

Abstract

Stochastic Gradient Descent (SGD) is a popular tool in training large-scale machine learning models. Its performance, however, is highly variable, depending crucially on the choice of the step sizes. Accordingly, a variety of strategies for tuning the step sizes have been proposed, ranging from coordinate-wise approaches (a.k.a. ``adaptive'' step sizes) to sophisticated heuristics to change the step size in each iteration. In this paper, we study two step size schedules whose power has been repeatedly confirmed in practice: the exponential and the cosine step sizes. For the first time, we provide theoretical support for them proving convergence rates for smooth non-convex functions, with and without the Polyak-Łojasiewicz (PL) condition. Moreover, we show the surprising property that these two strategies are \emph{adaptive} to the noise level in the stochastic gradients of PL functions. That is, contrary to polynomial step sizes, they achieve almost optimal performance without needing to know the noise level nor tuning their hyperparameters based on it. Finally, we conduct a fair and comprehensive empirical evaluation of real-world datasets with deep learning architectures. Results show that, even if only requiring at most two hyperparameters to tune, these two strategies best or match the performance of various finely-tuned state-of-the-art strategies.

References in corpus (22)

• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• A Simple Framework for Contrastive Learning of Visual Representations
• ADADELTA: An Adaptive Learning Rate Method
• On the Convergence of Adam and Beyond
• DARTS: Differentiable Architecture Search
• A Convergence Theory for Deep Learning via Over-Parameterization
• AdaGrad stepsizes: Sharp convergence over nonconvex landscapes
• Large Batch Optimization for Deep Learning: Training BERT in 76 minutes
• Bag of Freebies for Training Object Detection Neural Networks
• Adaptive Bound Optimization for Online Convex Optimization
• Online Learning: A Modern Introduction Using Convex Optimization
• Adversarial AutoAugment
• Closing the Generalization Gap of Adaptive Gradient Methods in Training Deep Neural Networks
• Stochastic Gradient Methods with Layer-wise Adaptive Moments for Training of Deep Networks
• Diverse Neural Network Learns True Target Functions
• Better Theory for SGD in the Nonconvex World
• How To Make the Gradients Small Stochastically: Even Faster Convex and Nonconvex SGD
• Online Adaptive Methods, Universality and Acceleration
• Deep Frank-Wolfe For Neural Network Optimization
• Graph Structure of Neural Networks
• PAGE: A Simple and Optimal Probabilistic Gradient Estimator for Nonconvex Optimization
• A High Probability Analysis of Adaptive SGD with Momentum

Cited by in corpus (6)

• Towards Understanding Label Smoothing
• WeMix: How to Better Utilize Data Augmentation
• On the Convergence of Stochastic Gradient Descent with Bandwidth-based Step Size
• On the Convergence of Step Decay Step-Size for Stochastic Optimization
• SMG: A Shuffling Gradient-Based Method with Momentum
• A Theoretical Analysis of Learning with Noisily Labeled Data