Meta-Learning Biologically Plausible Plasticity Rules with Random Feedback Pathways
Code to accompany the paper.
Modeling Neural Circuits Made Simple
Python notebooks to accompany the textbook. Code available on github and Google Colab.
RNNTools
Tools for rate-based and spike-based comp-neuro style RNN models in PyTorch. Under development.
Torch2PC
PyTorch software for using predictive coding algorithms to train PyTorch models.
Predictive Coding Vs BackProp.
Python code to compare predictive coding and backpropagation on image classification benchmarks.
To accompany the article On the relationship between predictive coding, backpropagation, and biological neuronal networks.
Simulations for evaluating the extent to which homeostatic plasticity learns to compute prediction errors.
Python code to simulate spiking and firing rate model networks with homeostatic plasticity.
To accompany the article Evaluating the extent to which homeostatic plasticity learns to compute prediction errors in unstructured neuronal networks.
FORCE, RMHL, and SUPERTREX on drawing tasks.
Matlab code to run the reservoir computing algorithms FORCE, RMHL, and SUPERTREX on three drawing tasks.
To accompany the article A model of reward-modulated motor learning with parallel cortical and basal ganglia pathways.
Correlated variability and asynchrony in heterogeneous and spatially extended spiking networks.
C and Matlab code to simulate large networks of EIF model neurons in correlated and asynchronous states.
To accompany the article The spatial structure of correlated neuronal variability.
Spiking statistics for the adaptive exponential integrate-and-fire (AdEx or AdEIF) neuron model driven by stochastic inputs.
C and Matlab code to efficiently compute the spiking statistics of an AdEx neuron model driven by stochastic input.
To accompany the article A diffusion approximation and numerical methods for adaptive neuron models with stochastic inputs.
Large spatially extended networks of LIF neurons.
C and Matlab code to simulate large networks of LIF neurons with spatially dependent recurrent connection probabilities. In one version of the code, random connections are computed on the fly by storing seeds for a random number generator. While this introduces some overhead in computation time, it avoids the need for storing a full connection matrix so that memory usage grows linearly instead of quadratically with network size. This permits the simulation of densely connected networks with millions of neurons.
To accompany the article Balanced networks of spiking neurons with spatially dependent recurrent connections.
Spiking statistics for the dLIF neuron model.
Matlab code for computing the spiking statistics of the discrete LIF neuron model.
To accompany the article Mechanisms that modulate the transfer of spiking correlations.
Nonautonomous Differential Equations.
Animations and Mathematica code to visualize the solution of some unstable linear nonautonomous differential equations with constant negative eigenvalues.
To accompany the article Unstable solutions of nonautonomous linear differential equations.
Code to accompany the paper.
Modeling Neural Circuits Made Simple
Python notebooks to accompany the textbook. Code available on github and Google Colab.
RNNTools
Tools for rate-based and spike-based comp-neuro style RNN models in PyTorch. Under development.
Torch2PC
PyTorch software for using predictive coding algorithms to train PyTorch models.
Predictive Coding Vs BackProp.
Python code to compare predictive coding and backpropagation on image classification benchmarks.
To accompany the article On the relationship between predictive coding, backpropagation, and biological neuronal networks.
Simulations for evaluating the extent to which homeostatic plasticity learns to compute prediction errors.
Python code to simulate spiking and firing rate model networks with homeostatic plasticity.
To accompany the article Evaluating the extent to which homeostatic plasticity learns to compute prediction errors in unstructured neuronal networks.
FORCE, RMHL, and SUPERTREX on drawing tasks.
Matlab code to run the reservoir computing algorithms FORCE, RMHL, and SUPERTREX on three drawing tasks.
To accompany the article A model of reward-modulated motor learning with parallel cortical and basal ganglia pathways.
Correlated variability and asynchrony in heterogeneous and spatially extended spiking networks.
C and Matlab code to simulate large networks of EIF model neurons in correlated and asynchronous states.
To accompany the article The spatial structure of correlated neuronal variability.
Spiking statistics for the adaptive exponential integrate-and-fire (AdEx or AdEIF) neuron model driven by stochastic inputs.
C and Matlab code to efficiently compute the spiking statistics of an AdEx neuron model driven by stochastic input.
To accompany the article A diffusion approximation and numerical methods for adaptive neuron models with stochastic inputs.
Large spatially extended networks of LIF neurons.
C and Matlab code to simulate large networks of LIF neurons with spatially dependent recurrent connection probabilities. In one version of the code, random connections are computed on the fly by storing seeds for a random number generator. While this introduces some overhead in computation time, it avoids the need for storing a full connection matrix so that memory usage grows linearly instead of quadratically with network size. This permits the simulation of densely connected networks with millions of neurons.
To accompany the article Balanced networks of spiking neurons with spatially dependent recurrent connections.
Spiking statistics for the dLIF neuron model.
Matlab code for computing the spiking statistics of the discrete LIF neuron model.
To accompany the article Mechanisms that modulate the transfer of spiking correlations.
Nonautonomous Differential Equations.
Animations and Mathematica code to visualize the solution of some unstable linear nonautonomous differential equations with constant negative eigenvalues.
To accompany the article Unstable solutions of nonautonomous linear differential equations.