Computer Science > Machine Learning
arXiv:2210.01117 (cs)
[Submitted on 3 Oct 2022 (v1), last revised 23 Mar 2023 (this version, v2)]
Title:Omnigrok: Grokking Beyond Algorithmic Data
View a PDF of the paper titled Omnigrok: Grokking Beyond Algorithmic Data, by Ziming Liu and 2 other authors
View PDFAbstract:Grokking, the unusual phenomenon for algorithmic datasets where generalization happens long after overfitting the training data, has remained elusive. We aim to understand grokking by analyzing the loss landscapes of neural networks, identifying the mismatch between training and test losses as the cause for grokking. We refer to this as the "LU mechanism" because training and test losses (against model weight norm) typically resemble "L" and "U", respectively. This simple mechanism can nicely explain many aspects of grokking: data size dependence, weight decay dependence, the emergence of representations, etc. Guided by the intuitive picture, we are able to induce grokking on tasks involving images, language and molecules. In the reverse direction, we are able to eliminate grokking for algorithmic datasets. We attribute the dramatic nature of grokking for algorithmic datasets to representation learning.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Data Analysis, Statistics and Probability (physics.data-an); Methodology (stat.ME); Machine Learning (stat.ML) |
| Cite as: | arXiv:2210.01117 [cs.LG] |
| (orarXiv:2210.01117v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2210.01117 arXiv-issued DOI via DataCite |
Submission history
From: Ziming Liu [view email][v1] Mon, 3 Oct 2022 17:58:04 UTC (6,110 KB)
[v2] Thu, 23 Mar 2023 13:42:27 UTC (6,431 KB)
Full-text links:
Access Paper:
- View PDF
- TeX Source
- Other Formats
View a PDF of the paper titled Omnigrok: Grokking Beyond Algorithmic Data, by Ziming Liu and 2 other authors
Current browse context:
cs.LG
References & Citations
Bookmark
Bibliographic and Citation Tools
Bibliographic Explorer(What is the Explorer?)
Connected Papers(What is Connected Papers?)
Litmaps(What is Litmaps?)
scite Smart Citations(What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv(What is alphaXiv?)
CatalyzeX Code Finder for Papers(What is CatalyzeX?)
DagsHub(What is DagsHub?)
Gotit.pub(What is GotitPub?)
Hugging Face(What is Huggingface?)
Papers with Code(What is Papers with Code?)
ScienceCast(What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower(What are Influence Flowers?)
CORE Recommender(What is CORE?)
IArxiv Recommender(What is IArxiv?)
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.