close this message
arXiv smileybones icon

Giving Week!

Show your support for Open Science by donating to arXiv during Giving Week, April 25th-29th.

DONATE
Skip to main content
Cornell University
We gratefully acknowledge support from
the Simons Foundation and member institutions.
arxiv logo > cs > arXiv:2110.06990

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Machine Learning

arXiv:2110.06990 (cs)
[Submitted on 13 Oct 2021 (v1), last revised 18 Oct 2021 (this version, v2)]

Title:Scaling Laws for the Few-Shot Adaptation of Pre-trained Image Classifiers

Authors:Gabriele Prato, Simon Guiroy, Ethan Caballero, Irina Rish, Sarath Chandar
Download PDF
Abstract: Empirical science of neural scaling laws is a rapidly growing area of significant importance to the future of machine learning, particularly in the light of recent breakthroughs achieved by large-scale pre-trained models such as GPT-3, CLIP and DALL-e. Accurately predicting the neural network performance with increasing resources such as data, compute and model size provides a more comprehensive evaluation of different approaches across multiple scales, as opposed to traditional point-wise comparisons of fixed-size models on fixed-size benchmarks, and, most importantly, allows for focus on the best-scaling, and thus most promising in the future, approaches. In this work, we consider a challenging problem of few-shot learning in image classification, especially when the target data distribution in the few-shot phase is different from the source, training, data distribution, in a sense that it includes new image classes not encountered during training. Our current main goal is to investigate how the amount of pre-training data affects the few-shot generalization performance of standard image classifiers. Our key observations are that (1) such performance improvements are well-approximated by power laws (linear log-log plots) as the training set size increases, (2) this applies to both cases of target data coming from either the same or from a different domain (i.e., new classes) as the training data, and (3) few-shot performance on new classes converges at a faster rate than the standard classification performance on previously seen classes. Our findings shed new light on the relationship between scale and generalization.
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2110.06990 [cs.LG]
  (or arXiv:2110.06990v2 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2110.06990
arXiv-issued DOI via DataCite

Submission history

From: Gabriele Prato [view email]
[v1] Wed, 13 Oct 2021 19:07:01 UTC (2,173 KB)
[v2] Mon, 18 Oct 2021 18:55:31 UTC (2,173 KB)
Full-text links:

Download:

  • PDF
  • Other formats
(license)
Current browse context:
cs.LG
< prev   |   next >
new | recent | 2110
Change to browse by:
cs
cs.AI
cs.CV

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar

DBLP - CS Bibliography

listing | bibtex
Gabriele Prato
Simon Guiroy
Ethan Caballero
Irina Rish
Sarath Chandar
a export bibtex citation Loading...

Bibtex formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Mendeley logo Reddit logo ScienceWISE logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code and Data Associated with this Article

arXiv Links to Code & Data (What is Links to Code & Data?)

Demos

Replicate (What is Replicate?)

Recommenders and Search Tools

Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs and how to get involved.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
  • About
  • Help
  • contact arXivClick here to contact arXiv Contact
  • subscribe to arXiv mailingsClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack