Skip to content

All working papers

Working paper No. 005

Forgetting on purpose

Authors: HyperPrior ResearchLondon
Research: 02Reinforcement learningLearning by acting, in a world that moves while you learn.
Programme: P-02FiltrationSequence models for drifting data
Status: In preparation

Version: In preparation
Date: In preparation
Confidence: In preparation
Revisit: In preparation

Abstract

A model that learns from a stream has to keep deciding how much of the past still applies. We make the case for models that forget on purpose, and for measuring when and why they let go.

If you work on related questions, write to us.

Cite

HYPERPRIOR RESEARCH LTDNo. 005

HyperPrior Research. Forgetting on purpose. Working paper No. 005, in preparation, 2026. London. https://hyperprior.net/papers/forgetting-on-purpose/

@techreport{hyperprior005,
  author      = {{HyperPrior Research}},
  title       = {Forgetting on purpose},
  type        = {Working paper},
  number      = {005},
  institution = {HyperPrior Research},
  address     = {London},
  year        = {2026},
  note        = {In preparation},
  url         = {https://hyperprior.net/papers/forgetting-on-purpose/}
}