Quantitative Biology > Neurons and Cognition
[Submitted on 30 Nov 2012 (this version), latest version 16 May 2013 (v3)]
Title:Encoding binary patterns of neural activity in networks of threshold-linear neurons
View PDFAbstract:Networks of neurons in the brain encode memories via their synaptic connections. Despite receiving considerable attention, the precise relationship between network connectivity and encoded activity patterns is still poorly understood. In particular, given a prescribed list of binary patterns, it is not generally known how to arrange the connectivity of a network so that exactly those patterns are encoded, while avoiding unwanted "spurious" states. Here we consider this problem for networks of threshold-linear neurons. We introduce a simple encoding rule that selectively turns "on" synapses between neurons that co-appear in one or more patterns. The synapses are binary, in the sense of having only two states ("on" or "off"), but also graded, with heterogeneous weights drawn from an underlying synaptic strength matrix S. Our main results provide necessary and sufficient conditions on S guaranteeing that prescribed patterns can be encoded, while maintaining tight control over spurious states. As an application, we construct networks that encode hippocampal place field codes nearly exactly. We suggest that, in this context, spurious states can be advantageous, allowing neural codes to be accurately encoded from a highly undersampled set of patterns. To obtain our results, we use ideas from convex and distance geometry, such as Helly's Theorem and Cayley-Menger determinants, revealing a novel connection between these areas of mathematics and coding properties of neural networks.
Submission history
From: Carina Curto [view email][v1] Fri, 30 Nov 2012 22:43:11 UTC (76 KB)
[v2] Fri, 5 Apr 2013 22:45:57 UTC (121 KB)
[v3] Thu, 16 May 2013 20:44:24 UTC (121 KB)
Current browse context:
q-bio.NC
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
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.