https://arxiv.org/abs/2106.06981

close this message

Donate to arXiv

Please join the Simons Foundation and our generous member
organizations in supporting arXiv during our giving campaign
September 23-27. 100% of your contribution will fund improvements and
new initiatives to benefit arXiv's global scientific community.

DONATE

[secure site, no need to create account]

Skip to main content
Cornell University
We gratefully acknowledge support from
the Simons Foundation and member institutions.
 
arXiv.org > cs > arXiv:2106.06981
[                    ]

Help | Advanced Search

[All fields        ]
Search
arXiv
Cornell University Logo
[                    ] GO
quick links

  * Login
  * Help Pages
  * About

Computer Science > Machine Learning

arXiv:2106.06981 (cs)
[Submitted on 13 Jun 2021]

Title:Thinking Like Transformers

Authors:Gail Weiss, Yoav Goldberg, Eran Yahav
Download PDF

    Abstract: What is the computational model behind a Transformer?
    Where recurrent neural networks have direct parallels in finite
    state machines, allowing clear discussion and thought around
    architecture variants or trained models, Transformers have no
    such familiar parallel. In this paper we aim to change that,
    proposing a computational model for the transformer-encoder in
    the form of a programming language. We map the basic components
    of a transformer-encoder -- attention and feed-forward
    computation -- into simple primitives, around which we form a
    programming language: the Restricted Access Sequence Processing
    Language (RASP). We show how RASP can be used to program
    solutions to tasks that could conceivably be learned by a
    Transformer, and how a Transformer can be trained to mimic a RASP
    solution. In particular, we provide RASP programs for histograms,
    sorting, and Dyck-languages. We further use our model to relate
    their difficulty in terms of the number of required layers and
    attention heads: analyzing a RASP program implies a maximum
    number of heads and layers necessary to encode a task in a
    transformer. Finally, we see how insights gained from our
    abstraction might be used to explain phenomena seen in recent
    works.

Comments: ICML 2021
Subjects: Machine Learning (cs.LG); Computation and Language (cs.CL)
Cite as:  arXiv:2106.06981 [cs.LG]
          (or arXiv:2106.06981v1 [cs.LG] for this version)

Submission history

From: Gail Weiss [view email]
[v1] Sun, 13 Jun 2021 13:04:46 UTC (3,616 KB)
Full-text links:

Download:

  * PDF
  * Other formats

(license)
Current browse context:
cs.LG
< prev   |   next >
new | recent | 2106
Change to browse by:
cs
cs.CL

References & Citations

  * NASA ADS
  * Google Scholar
  * Semantic Scholar

a export bibtex citation Loading...

Bibtex formatted citation

x
[loading...          ]
Data provided by:

Bookmark

BibSonomy logo Mendeley logo Reddit logo ScienceWISE logo
(*) Bibliographic Tools

Bibliographic and Citation Tools

[ ] Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
( ) Code & Data

Code and Data Associated with this Article

[ ] arXiv Links to Code Toggle
arXiv Links to Code & Data (What is Links to Code & Data?)
( ) Related Papers

Recommenders and Search Tools

[ ] Connected Papers Toggle
Connected Papers (What is Connected Papers?)
[ ] Core recommender toggle
CORE Recommender (What is CORE?)
( ) About arXivLabs

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

  * Click here to contact arXiv Contact
  * Click here to subscribe Subscribe

  * Copyright
  * Privacy Policy

  * Web Accessibility Assistance
  * arXiv Operational Status
    Get status notifications via email or slack