Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Short-term memory in orthogonal neural networks.

Olivia L White1, Daniel D Lee, Haim Sompolinsky

  • 1Harvard University, Cambridge, Massachusetts 02138, USA.

Physical Review Letters
|April 20, 2004
PubMed
Summary

Linear recurrent networks can store long temporal sequences, with memory capacity scaling with network size. This finding is crucial for understanding temporal data processing in artificial neural networks.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Uncovering Sex Differences in the Drosophila Ventral Nerve Cord Through Connectome Alignment.

bioRxiv : the preprint server for biology·2026
Same author

Oral [<sup>18</sup>F]-Fluoro-Thia-Heptadecanoic Acid Positron Emission Tomography Reveals Mesenteric-to-Central Lymphatic Flow.

Gastro hep advances·2026
Same author

Unraveling the geometry of visual relational reasoning.

Scientific reports·2026
Same author

The redundant protein synthesis gene Aimp1 challenges the canonical inverse relationship between translation and autophagy.

Communications biology·2026
Same author

A Gut-Restricted Liver X Receptor Agonist Ameliorates Liver Injury in Experimental Short Bowel Syndrome.

Gastroenterology·2026
Same author

Lymphatic disruption drives lung transplant fibrosis through interleukin-1-mediated hyaluronan accumulation.

Science translational medicine·2026

Area of Science:

  • Computational neuroscience
  • Machine learning
  • Dynamical systems

Background:

  • Recurrent neural networks (RNNs) are essential for processing sequential data.
  • Understanding the limits of temporal information storage in RNNs is a key challenge.

Purpose of the Study:

  • To investigate the temporal memory capacity of linear recurrent networks.
  • To determine how memory capacity scales with network size.

Main Methods:

  • Analysis of discrete-time linear recurrent networks.
  • Calculation of temporal memory capacity for specific connectivity types (distributed shift register, random orthogonal).

Main Results:

  • Temporal memory capacity was calculated for different network configurations.
  • The memory capacity was shown to scale with the system size of the networks.

Conclusions:

  • Linear recurrent networks possess a quantifiable temporal memory capacity.
  • Network size is a critical factor influencing the ability to store long temporal sequences.

Related Experiment Videos