Computing over in-vitro predictive coding neural cultures

While artificial neural networks are gaining in popularity, they still fall behind biological neural substrates in terms of energy efficiency and performance on certain computational tasks. However, the mechanisms by which neurons operate are still not well understood, making it difficult to harness their computational power for arbitrary tasks. Predictive coding is an influential theory of learning and inference within neuroscience, positing that neural systems adapt to best predict sensory input across time, thus representing the sensory distribution within an internal generative model encoded through synaptic connections. However, previous work on predictive coding has been limited to modeling relations between higher-level units of the brain. Here, we present a biologically plausible model of predictive coding generalized to arbitrary topologies of in-vitro cultures. In addition, we present a novel framework to harness neural cultures that implement predictive coding for computational tasks.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Jain, Shrusti

Contributors dc:contributor

• Rauchwerger, Lawrence

• Predictive Coding
• Neuroscience
• Bnns

Statement dc:rights

• Copyright 2025 Shrusti Jain

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/129979

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F129979