Jump to content

Double descent

From Wikipedia, the free encyclopedia
An example of the double descent phenomenon in a two-layer neural network: When the ratio of parameters to data points is increased, the test error falls first, then rises, then falls again.[1] The vertical line marks the "interpolation threshold" boundary between the underparametrized regime (more data points than parameters) and the overparameterized regime (more parameters than data points).

In statistics and machine learning, double descent is the phenomenon where a statistical model with a small number of parameters and a model with an extremely large number of parameters have a small test error, but a model whose number of parameters is about the same as the number of data points used to train the model will have a large error.[2] This phenomenon has been considered surprising, as it contradicts assumptions about overfitting in classical machine learning.[1]

History

[edit]

Early observations of what would later be called double descent in specific models date back to 1989.[3][4]

The term "double descent" was coined by Belkin et. al.[5] in 2019,[1] when the phenomenon as a broader concept shared by many models gained popularity.[6][7] The latter development was prompted by a perceived contradiction between the conventional wisdom that too many parameters in the model result in a significant overfitting error (an extrapolation of bias-variance tradeoff),[8] and the empirical observations in the 2010s that some modern machine learning models tend to perform better with larger models.[5][9]

Theoretical models

[edit]

Double descent occurs in linear regression with isotropic Gaussian covariates and isotropic Gaussian noise.[10]

A model of double descent at the thermodynamic limit has been analyzed by the replica method, and the result has been confirmed numerically.[11]

Empirical examples

[edit]

The scaling behavior of double descent has been found to follow a broken neural scaling law[12] functional form.

References

[edit]
  1. ^ a b c Schaeffer, Rylan; Khona, Mikail; Robertson, Zachary; Boopathy, Akhilan; Pistunova, Kateryna; Rocks, Jason W.; Fiete, Ila Rani; Koyejo, Oluwasanmi (2023-03-24). "Double Descent Demystified: Identifying, Interpreting & Ablating the Sources of a Deep Learning Puzzle". arXiv:2303.14151v1 [cs.LG].
  2. ^ "Deep Double Descent". OpenAI. 2019-12-05. Retrieved 2022-08-12.
  3. ^ Vallet, F.; Cailton, J.-G.; Refregier, Ph (June 1989). "Linear and Nonlinear Extension of the Pseudo-Inverse Solution for Learning Boolean Functions". Europhysics Letters. 9 (4): 315. Bibcode:1989EL......9..315V. doi:10.1209/0295-5075/9/4/003. ISSN 0295-5075.
  4. ^ Loog, Marco; Viering, Tom; Mey, Alexander; Krijthe, Jesse H.; Tax, David M. J. (2020-05-19). "A brief prehistory of double descent". Proceedings of the National Academy of Sciences. 117 (20): 10625–10626. arXiv:2004.04328. Bibcode:2020PNAS..11710625L. doi:10.1073/pnas.2001875117. ISSN 0027-8424. PMC 7245109. PMID 32371495.
  5. ^ a b Belkin, Mikhail; Hsu, Daniel; Ma, Siyuan; Mandal, Soumik (2019-08-06). "Reconciling modern machine learning practice and the bias-variance trade-off". Proceedings of the National Academy of Sciences. 116 (32): 15849–15854. arXiv:1812.11118. doi:10.1073/pnas.1903070116. ISSN 0027-8424. PMC 6689936. PMID 31341078.
  6. ^ Spigler, Stefano; Geiger, Mario; d'Ascoli, Stéphane; Sagun, Levent; Biroli, Giulio; Wyart, Matthieu (2019-11-22). "A jamming transition from under- to over-parametrization affects loss landscape and generalization". Journal of Physics A: Mathematical and Theoretical. 52 (47): 474001. arXiv:1810.09665. doi:10.1088/1751-8121/ab4c8b. ISSN 1751-8113.
  7. ^ Viering, Tom; Loog, Marco (2023-06-01). "The Shape of Learning Curves: A Review". IEEE Transactions on Pattern Analysis and Machine Intelligence. 45 (6): 7799–7819. arXiv:2103.10948. doi:10.1109/TPAMI.2022.3220744. ISSN 0162-8828. PMID 36350870.
  8. ^ Geman, Stuart; Bienenstock, Élie; Doursat, René (1992). "Neural networks and the bias/variance dilemma" (PDF). Neural Computation. 4: 1–58. doi:10.1162/neco.1992.4.1.1. S2CID 14215320.
  9. ^ Preetum Nakkiran; Gal Kaplun; Yamini Bansal; Tristan Yang; Boaz Barak; Ilya Sutskever (29 December 2021). "Deep double descent: where bigger models and more data hurt". Journal of Statistical Mechanics: Theory and Experiment. 2021 (12). IOP Publishing Ltd and SISSA Medialab srl: 124003. arXiv:1912.02292. Bibcode:2021JSMTE2021l4003N. doi:10.1088/1742-5468/ac3a74. S2CID 207808916.
  10. ^ Nakkiran, Preetum (2019-12-16). "More Data Can Hurt for Linear Regression: Sample-wise Double Descent". arXiv:1912.07242v1 [stat.ML].
  11. ^ Advani, Madhu S.; Saxe, Andrew M.; Sompolinsky, Haim (2020-12-01). "High-dimensional dynamics of generalization error in neural networks". Neural Networks. 132: 428–446. doi:10.1016/j.neunet.2020.08.022. ISSN 0893-6080. PMC 7685244. PMID 33022471.
  12. ^ Caballero, Ethan; Gupta, Kshitij; Rish, Irina; Krueger, David (2022). "Broken Neural Scaling Laws". International Conference on Learning Representations (ICLR), 2023.

Further reading

[edit]
[edit]