Predictive Coding with Cortical Cell Types and Connectivity

Join us for a talk by Hannah Choi, Assistant Professor, School of Mathematics, Georgia Institute of Technology. This talk is part of the Kempner Seminar Series, a research-level seminar series that covers topics related to the basis of intelligence in natural and artificial systems.

The brain efficiently processes sensory information by generating an internal model of the environment, which is continuously updated through prediction errors and the suppression of expected information— a process known as predictive coding. While experimental evidence supports cortical implementation of predictive coding, it remains unclear how this process is shaped by connectivity across the cortical hierarchy, layers, and diverse neuronal subtypes. I will discuss recent work from my group that investigates cortical mechanisms underlying predictive coding.

By constructing a biologically grounded network model with realistic connectivity among cortical cell types, our work uncovers context-dependent functional connectivity that reflects the communication of predictions and prediction errors across the cortical circuits. Furthermore, by mapping algorithmic components of predictive coding onto neuronal subtypes, our study generates experimentally testable predictions about cell-type-specific responses to expected and unexpected stimuli. We show how different computational objectives such as prediction error minimization, energy efficiency, and reward maximization shape cell-type-specific responses to distinct forms of expectation violation, including contextual, absolute, and omission novelty. Our study thus integrates a multi-objective normative framework with a biological circuit model, offering a unified account of how predictive coding, energy efficiency, and reinforcement learning map onto cell-type-specific mechanisms.