The DOI matches its Crossref record; 32 of the 32 references it cites were found in a registry or an index.
Super-convergence: very fast training of neural networks using large learning rates
Open accessAbstract
In this paper, we describe a phenomenon, which we named “super-convergence”, where neural networks can be trained an order of magnitude faster than with standard training methods. The existence of super-convergence is relevant to understanding why deep networks generalize well. One of the key elements of super-convergence is training with one learning rate cycle and a large maximum learning rate. A insight that allows super-convergence training is that large learning rates regularize the training, hence requiring a reduction of all other forms of regularization in order to preserve an optimal regularization balance. We also derive a simplification of the Hessian Free optimization method to compute an estimate of the optimal learning rate. Experiments demonstrate super-convergence for Cifar-10/100, MNIST and Imagenet datasets, and resnet, wide-resnet, densenet, and inception architectures. In addition, we show that super-convergence provides a greater boost in performance relative to standard training when the amount of labeled training data is limited. The architectures and code to replicate the figures in this paper are available at github.com/lnsmith54/super-convergence.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
(2010) · Cited by 8,261
Bengio, Y., Louradour, J., Collobert, R., & Weston, J. (2009) · Proceedings of the 26th Annual International Conference on Machine Learning · doi:10.1145/1553374.1553380 · Cited by 5,366 · cited 1×
“Please note that the philosophy behind CLR is a combination of curriculum learning [Bengio et al., 2009] and simulated annealing [Aarts and Korst, 1988], both of which have a long history of use in deep learning.”
BackgroundADADELTA: An Adaptive Learning Rate Method
Zeiler, M. D. (2012) · arXiv · doi:10.48550/arxiv.1212.5701 · Cited by 5,322
SGDR: Stochastic Gradient Descent with Warm Restarts
Loshchilov, I., & Hutter, F. (2016) · arXiv · doi:10.48550/arxiv.1608.03983 · Cited by 5,309
(2013) · Cited by 3,740
Optimization Methods for Large-Scale Machine Learning
Bottou, L., Curtis, F. E., & Nocedal, J. (2018) · SIAM Review · doi:10.1137/16m1080173 · Cited by 3,128
Understanding deep learning requires rethinking generalization
Zhang, C., Bengio, S., Hardt, M., Recht, B., & Vinyals, O. (2016) · arXiv · doi:10.48550/arxiv.1611.03530 · Cited by 2,838
Cyclical Learning Rates for Training Neural Networks
Smith, L. N. (2017) · 2017 IEEE Winter Conference on Applications of Computer Vision (WACV) · doi:10.1109/wacv.2017.58 · Cited by 2,670
Self-citationAccurate, Large Minibatch SGD: Training ImageNet in 1 Hour
Goyal, P., Dollár, P., Girshick, R., Noordhuis, P., Wesolowski, L., Kyrola, A., Tulloch, A., Jia, Y., & He, K. (2017) · arXiv · doi:10.48550/arxiv.1706.02677 · Cited by 2,294
(2013) · Cited by 2,056
Stochastic Gradient Descent Tricks
Bottou, L. (2012) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-35289-8_25 · Cited by 1,929
*13 Simulated Annealing and Boltzmann Machines
(1993) · Neural Network Learning and Expert Systems · doi:10.7551/mitpress/4931.003.0017 · Cited by 1,637
A Closer Look at Memorization in Deep Networks
Arpit, D., Jastrzębski, S., Ballas, N., Krueger, D., Bengio, E., Kanwal, M. S., Maharaj, T., Fischer, A., Courville, A., Bengio, Y., et al. (2017) · arXiv · doi:10.48550/arxiv.1706.05394 · Cited by 1,307
Smith, L. N. (2018) · arXiv · doi:10.48550/arxiv.1803.09820 · Cited by 790
Self-citationDeep learning via Hessian-free optimization
Martens, J. (2010) · International Conference on Machine Learning · Cited by 732
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.1.1 · Cited by 676
Don't Decay the Learning Rate, Increase the Batch Size
Smith, S. L., Kindermans, P.-J., Ying, C., & Le, Q. V. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.00489 · Cited by 559
Noisy Networks for Exploration
Fortunato, M., Azar, M. G., Piot, B., Menick, J., Osband, I., Graves, A., Mnih, V., Munos, R., Hassabis, D., Pietquin, O., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1706.10295 · Cited by 543
On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima
Keskar, N. S., Mudigere, D., Nocedal, J., Smelyanskiy, M., & Tang, P. T. P. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1609.04836 · Cited by 525
Qualitatively characterizing neural network optimization problems
Goodfellow, I. J., Vinyals, O., & Saxe, A. M. (2014) · arXiv · doi:10.48550/arxiv.1412.6544 · Cited by 341
Schaul, T., Zhang, S., & LeCun, Y. (2012) · arXiv (Cornell University) · doi:10.48550/arxiv.1206.1106 · Cited by 266
Towards Understanding Generalization of Deep Learning: Perspective of Loss LandscapesPreprint
Wu, L., Zhu, Z.-X., & Weinan, E. (2017) · arXiv.org · doi:10.48550/arxiv.1706.10239 · Cited by 250
Three Factors Influencing Minima in SGD
Jastrzȩbski, S., Kenton, Z., Arpit, D., Ballas, N., Fischer, A., Bengio, Y., & Storkey, A. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.04623 · Cited by 217
Chaudhari, P., & Soatto, S. (2018) · 2018 Information Theory and Applications Workshop (ITA) · doi:10.1109/ita.2018.8503224 · Cited by 213
Generalization with Deep Learning: For Improvement on Sensing Capability
Chen, Z., Wu, M., & Li, X. (2020) · WORLD SCIENTIFIC eBooks · doi:10.1142/11784 · Cited by 179
No More Pesky Learning Rate Guessing Games.Preprint
Smith, L. N. (2015) · arXiv (Cornell University) · Cited by 154
Self-citationEntropy-SGD: Biasing Gradient Descent Into Wide ValleysPreprint
Chaudhari, P., Choromanska, A., Soatto, S., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1611.01838 · Cited by 104
An empirical analysis of the optimization of deep network loss surfacesPreprint
Im, D. J., Tao, M. W., & Branson, K. M. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1612.04010 · Cited by 78
ADASECANT: Robust Adaptive Secant Method for Stochastic GradientPreprint
Gülçehre, Ç., Moczulski, M., & Bengio, Y. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1412.7419 · Cited by 21
Hoffer, E., Hubara, I., & Soudry, D. (2017) · arXiv · doi:10.48550/arxiv.1705.08741
Cited by
AI Feynman: A physics-inspired method for symbolic regression
Udrescu, S.-M., & Tegmark, M. (2020) · Science Advances · doi:10.1126/sciadv.aay2631 · Cited by 1,080
Smith, L. N. (2018) · arXiv · doi:10.48550/arxiv.1803.09820 · Cited by 790
Self-citationAdaBins: Depth Estimation Using Adaptive Bins
Farooq Bhat, S., Alhashim, I., & Wonka, P. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00400 · Cited by 726
Fast is better than free: Revisiting adversarial training
Wong, E., Rice, L., & Kolter, J. Z. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2001.03994 · Cited by 544
RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching
Lipson, L., Teed, Z., & Deng, J. (2021) · 2021 International Conference on 3D Vision (3DV) · doi:10.1109/3dv53792.2021.00032 · Cited by 536
Metadata
- Authors
- Smith, L. N., & Topin, N.
- Venue
- Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications
- Publisher
- SPIE
- Published
- 2019-05-10
- Pages
- 36
- OpenAlex
- W2749581528
- Cited by
- 1,335
- References
- 32
- FWCI
- 43.37
- Type
- conference paper
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.