How to decay your learning rate

Complex learning rate schedules have become an integral part of deep learning. We find empirically that common fine-tuned schedules decay the learning rate after the weight norm bounces. This leads to the proposal of ABEL: an automatic scheduler which decays the learning rate by keeping track of the weight norm. ABEL’s performance matches that of tuned schedules and is more robust with respect to its parameters. Through extensive experiments in vision, NLP, and RL, we show that if the weight norm does not bounce, we can simplify schedules even further with no loss in performance. In such cases, a complex schedule has similar performance to a constant learning rate with a decay at the end of training.

[1]  Preetum Nakkiran.  Learning Rate Annealing Can Provably Help Generalization, Even for Convex Problems , 2020, ArXiv.

[2]  Sho Yaida,et al.  Fluctuation-dissipation relations for stochastic gradient descent , 2018, ICLR.

[3]  Aitor Lewkowycz,et al.  On the training dynamics of deep networks with L2 regularization , 2020, NeurIPS.

[4]  Twan van Laarhoven,et al.  L2 Regularization versus Batch and Weight Normalization , 2017, ArXiv.

[5]  Mark Chen,et al.  Language Models are Few-Shot Learners , 2020, NeurIPS.

[6]  Naman Agarwal,et al.  Disentangling Adaptive Gradient Methods from Learning Rates , 2020, ArXiv.

[7]  L. Eon Bottou.  Online Learning and Stochastic Approximations , 1998 .

[8]  Sanjeev Arora,et al.  Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate , 2020, NeurIPS.

[9]  Quoc V. Le,et al.  EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks , 2019, ICML.

[10]  Hossein Mobahi,et al.  Sharpness-Aware Minimization for Efficiently Improving Generalization , 2020, ArXiv.

[11]  Alec Radford,et al.  Scaling Laws for Neural Language Models , 2020, ArXiv.

[12]  Sanjeev Arora,et al.  An Exponential Learning Rate Schedule for Deep Learning , 2020, ICLR.

[13]  Kaifeng Lyu,et al.  Gradient Descent Maximizes the Margin of Homogeneous Neural Networks , 2019, ICLR.

[14]  Colin Wei,et al.  Towards Explaining the Regularization Effect of Initial Large Learning Rate in Training Neural Networks , 2019, NeurIPS.

[15]  Misha Denil,et al.  Learned Optimizers that Scale and Generalize , 2017, ICML.

[16]  Ryan P. Adams,et al.  Gradient-based Hyperparameter Optimization through Reversible Learning , 2015, ICML.

[17]  Jian Sun,et al.  Spherical Motion Dynamics: Learning Dynamics of Neural Network with Normalization, Weight Decay, and SGD , 2020 .

[18]  Ming-Wei Chang,et al.  BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[19]  Jascha Sohl-Dickstein,et al.  Measuring the Effects of Data Parallelism on Neural Network Training , 2018, J. Mach. Learn. Res..