infinityabundance/learning-under-irreversibility
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
A structural account of learning as commitment revision constrained by irreversible time.
5 commits 1 branch 0 tags 218 KiB
Find a file
Exact
riaandebeer 6dce33dee3
 Add files via upload
2026-02-11 05:01:18 +00:00
CITATION.cff Create CITATION.cff 2026-02-11 05:00:55 +00:00
Learning as Constrained Commitment Revision.pdf Add files via upload 2026-02-11 05:01:18 +00:00
LICENSE Create LICENSE 2026-02-11 05:00:35 +00:00
README.md Update README.md 2026-02-11 05:00:15 +00:00

README.md

Learning Under Irreversibility

DOI

This repository provides a stable mirror of the paper:

Learning as Constrained Commitment Revision Under Irreversible Time
by Riaan de Beer (2026).

The canonical, citable version of this work is published on Zenodo and should be referenced using the DOI above.

Citation

de Beer, R. (2026).
Learning as Constrained Commitment Revision Under Irreversible Time (v1.0).
Zenodo. https://doi.org/10.5281/zenodo.18604349

About

This work presents a structural account of learning in systems that persist under irreversible time. Learning is formalized as commitment revision constrained by preservation of existing binding structure. Updates that preserve commitments maintain identity and coherence, while updates that violate binding lead to fragmentation and instability.

Catastrophic forgetting is reinterpreted as commitment violation rather than information loss, and continual learning is shown to succeed only when updates remain compatible with accumulated commitments. The paper identifies architectural requirements for identity-preserving learning without assuming reversible modification or unrestricted overwrite.

Status

  • This repository exists solely to provide redundancy and long-term access.
  • The paper is complete and not under active development.
  • No revisions or extensions are planned for this version.

License

This work is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
See the LICENSE file for details.