HDPC_code

This repository contains simulation code for Wang, Audette, Schneider, and Aljadeff, "Desegregation of neuronal predictive processing," bioRxiv (2024).

The code studies predictive-processing neural networks under steady-state, pulse-input, sparse-connectivity, hierarchical, and inhibitory-neuron conditions. Most files are MATLAB scripts; the inhibitory-neuron example is a Wolfram Mathematica notebook that loads a Wolfram package from this repository.

Requirements

• MATLAB, tested with recent MATLAB releases.
• MATLAB Statistics and Machine Learning Toolbox for mvnrnd.
• Wolfram Mathematica 13 or later for Inhibitory_response.nb.

Files

File	Language	Purpose	Main output
SS_network.m	MATLAB	Simulates steady-state responses in the base predictive-processing network.	ss_response.mat with hf and rf.
SS_TwoPair.m	MATLAB	Simulates four match and mismatch conditions for two stimulus pairs. These responses are used to characterize cell types for Fig. 4c.	PairedRep_*.mat with mmx1, m1, mmy2, and m2.
Pulse_Input_Network.m	MATLAB	Simulates time-dependent responses to pulse-like x and y inputs, plus x-only and y-only controls.	TDcode_longtxy_*.mat with time vectors, voltage responses, weights, and parameters.
SS_hirachical_net.m	MATLAB	Simulates steady-state responses in the hierarchical predictive-processing network. The filename keeps the original spelling used by the project.	Hierachical_response.mat with hf and rf.
Sparse_network.m	MATLAB	Simulates time-dependent responses in a sparsified, asymmetric predictive-processing network.	Sparse_NN.mat with t1, h1, r1, weights, and parameters.
IneuronRep.m	Wolfram Language	Defines helper functions for inhibitory-neuron response calculations.	Functions loaded by Inhibitory_response.nb.
Inhibitory_response.nb	Mathematica notebook	Runs inhibitory-neuron response calculations by loading IneuronRep.m.	Notebook variables hImis, hImisy, and hIfull.

Basic Usage

Run each MATLAB script from the repository root so output files are written next to the code:

SS_network
SS_TwoPair
Pulse_Input_Network
Sparse_network
SS_hirachical_net

To run the inhibitory-neuron notebook, open Inhibitory_response.nb in Mathematica from the repository root. The notebook uses:

Get["IneuronRep.m"]

so IneuronRep.m must remain in the same directory.

Changing Stimulus Conditions

Stimulus conditions are controlled by the vectors x and y in the MATLAB scripts. They should stay as p-by-1 column vectors because the model multiplies them by the sampled weight matrices w and v.

Common parameters near the top of each script include:

• nsize or size: number of neurons.
• p: number of stimulus pairs.
• mu: learning or stimulus-correlation parameter.
• b: model gain parameter.
• theta: firing threshold.
• tf, tpre, txy, and delta_t: simulation timing parameters.

Notes

• Scripts use random weight vectors, so repeated runs produce different .mat files unless you set a random seed with rng(...) before sampling.
• The scripts are written as direct, editable simulation files. Change parameters near the top of a file, run the script, and inspect the saved .mat output.
• Generated .mat files are analysis outputs and are not required to run the source code.