A genetic algorithm for self-supervised models of oscillatory neurodynamics.

Code ↔ Paper

7 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 7 matches

[1] § Materials and methods ↔ misc/Jaxley_Mechanisms.ipynb, lines 473–522 · score 0.65 · stochastic delta rule, model parameters, 0–1, logic, genetic, supervised
[2] § Materials and methods ↔ gsdr/optimizers.py, lines 222–352 · score 0.56 · deselection threshold, MCDP factors, lambda, exploration, stochastic, optimization
[3] § Materials and methods ↔ misc/Jaxley_Mechanisms.ipynb, lines 524–564 · score 0.55 · exploration factor, deselection threshold, reverts, optimization, training, loss
[4] § Results ↔ Biophys_SX.ipynb, lines 3117–3178 · score 0.55 · beta band power, gamma band power, synaptic weight, connectivity, model
[5] § Materials and methods ↔ misc/Jaxley_Mechanisms.ipynb, lines 473–522 · score 0.54 · Genetic Stochastic Delta, stochastic delta rule, model parameter, deselects, dynamics, optimization
[6] § Materials and methods ↔ gsdr/analysis.py, lines 85–96 · score 0.53 · Mutual correlation dependent, plasticity
[7] § Materials and methods ↔ misc/Jaxley_Mechanisms.ipynb, lines 49–139 · score 0.52 · pre synaptic, post synaptic, tau, activity, spike, model

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

Jupyter notebook · 638 lines · 21 KB · no license · 4 matches

# %% [markdown]
# <a href="https://colab.research.google.com/github/HNXJ/GSDR/blob/main/Jaxley_Mechanisms.ipynb" target="_parent"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
# %% [markdown]
# Jaxley Biophysical model mechanisms
#
# This notebook contains definitions of biophysical mechanisms to be used in Jaxley https://jaxley.readthedocs.io/en/
#
# @HNyXJ (Assisted by Vanderbilt's amplify AI and Google's Gemini)
#
# Mechanisms included or to be implemented in this notebook :
#
# GABAa
#
# GABAb
#
# AMPA
#
# NMDA
#
# ACh
#
# D1
#
# D2
#
# 5HT
#
# %%
%pip install jaxley
from jax import config
config.update("jax_enable_x64", True)
config.update("jax_platform_name", "cpu")
import matplotlib.pyplot as plt
import numpy as np
import jax
import jax.numpy as jnp
from jax import jit, vmap, value_and_grad
import jaxley as jx
from jaxley.channels import Leak, HH
from jaxley.synapses import IonotropicSynapse
from jaxley.connect import fully_connect
# %% [markdown]
# # Neuronal mechanisms
# %% [markdown]
# ## GABAa
# %%
import jax.numpy as jnp
from jaxley.synapses import Synapse
class GradedGABAa(Synapse):
"""
A graded (non-spiking) GABAa synapse model based on high-threshold
graded transmission dynamics.
Unlike standard exponential synapses which are triggered by spike events,
this mechanism's gating variable 's' is continuously driven by the
presynaptic voltage via a hyperbolic tangent transfer function.
Dynamics:
ds/dt = -s/tauD + (1/2) * (1 + tanh((V_pre - V_th) / slope)) * ((1-s)/tauR)
I_syn = gGABAa * s * (V_post - EGABAa)
Parameters:
gGABAa (float): Peak synaptic conductance [uS or mS/cm^2 depending on context].
EGABAa (float): Reversal potential [mV]. Default: -80.0.
tauD (float): Decay time constant [ms]. Default: 10.0.
tauR (float): Rise time constant [ms]. Default: 0.2.
V_th (float): Half-activation voltage [mV]. Default: 0.0 (centered tanh).
slope (float): Sensitivity of the activation curve [mV]. Default: 10.0.
References:
1. Golowasch, J., Casey, M., Abbott, L. F., & Marder, E. (1999).
Network stability from neuronal insight. Journal of Neurobiology, 41(3), 331-348.
2. Prinz, A. A., Bucher, D., & Marder, E. (2004).
Similar network activity from disparate circuit parameters.
Nature Neuroscience, 7(12), 1345-1352.
"""
def __init__(self, name: str = "GradedGABAa"):
super().__init__(name)
# Parameter definitions matching DynaSim defaults
self.synapse_params = {
"gGABAa": 0.25,
"EGABAa": -80.0,
"tauD": 10.0,
"tauR": 0.2,
"slope": 10.0, # Denominator inside the tanh
"V_th": 0.0 # Midpoint of the tanh (implicit 0 in DynaSim)
}
# Initial Condition (IC) matching DynaSim IC=[0.1]
self.synapse_states = {"s": 0.1}
def update_states(self, states, dt, pre_v, post_v, params):
"""
Updates the gating variable 's' based on PRE-synaptic voltage.
"""
s = states["s"]
tauD = params["tauD"]
tauR = params["tauR"]
slope = params["slope"]
v_th = params["V_th"]
# The transfer function (Activation)
# Corresponds to DynaSim: 1/2 * (1 + tanh(X_pre / 10))
# We added V_th to make it more robust, but it defaults to 0.
activation = 0.5 * (1 + jnp.tanh((pre_v - v_th) / slope))
# Differential equation:
# s' = -s/tauD + activation * (1-s)/tauR
d_s = (-s / tauD) + activation * ((1 - s) / tauR)
# Forward Euler integration
new_s = s + d_s * dt
return {"s": new_s}
def compute_current(self, states, pre_v, post_v, params):
"""
Calculates the synaptic current flowing into the POST-synaptic cell.
"""
s = states["s"]
g = params["gGABAa"]
e_rev = params["EGABAa"]
# Ohm's Law for the synapse
current = g * s * (post_v - e_rev)
return current
# %%
import jaxley as jx
# Instantiate the channel
gaba_mech = GABAa()
# Create a cell and add the mechanism
cell = jx.Cell()
cell.insert(gaba_mech)
# If you need to modify parameters specifically for this cell:
cell.GABAa.gGABAa = 0.5 # Override default conductance
# %%
net.delete_recordings()
net.cell(0).branch(0).loc(0.0).record()
net.cell(1).branch(0).loc(0.0).record()
net.cell(2).branch(0).loc(0.0).record()
# %%
inputs = jnp.asarray(np.random.rand(100, 2))
labels = jnp.asarray((inputs[:, 0] + inputs[:, 1]) > 1.0)
# %%
fig, ax = plt.subplots(1, 1, figsize=(3, 2))
_ = ax.scatter(inputs[labels, 0], inputs[labels, 1])
_ = ax.scatter(inputs[~labels, 0], inputs[~labels, 1])
# %%
labels = labels.astype(float)
# net.edges
# %%
net.delete_trainables()
net.make_trainable("radius")
net.cell("all").branch("all").loc("all").make_trainable("Leak_gLeak")
net.IonotropicSynapse.edge("all").make_trainable("IonotropicSynapse_gS")
# %% [markdown]
#
#
# * II
#
# %%
params = net.get_parameters()
s = jx.integrate(net, params=params, t_max=5.0)
# %%
def simulate(params, inputs):
currents = jx.datapoint_to_step_currents(i_delay=10.0, i_dur=40.0, i_amp=10*inputs, delta_t=0.025, t_max=100.0)
data_stimuli = None
data_stimuli = net.cell(0).branch(2).loc(1.0).data_stimulate(currents[0], data_stimuli=data_stimuli)
data_stimuli = net.cell(1).branch(2).loc(1.0).data_stimulate(currents[1], data_stimuli=data_stimuli)
# data_stimuli = net.cell(3).branch(2).loc(1.0).data_stimulate(currents[1], data_stimuli=data_stimuli)
# data_stimuli = net.cell(4).branch(2).loc(1.0).data_stimulate(currents[1], data_stimuli=data_stimuli)
return jx.integrate(net, params=params, data_stimuli=data_stimuli, delta_t=0.025)
batched_simulate = vmap(simulate, in_axes=(None, 0))
# %%
traces = batched_simulate(params, inputs[:4])
fig, ax = plt.subplots(1, 1, figsize=(4, 2))
_ = ax.plot(traces[:, 2, :].T)
# %%
def rasterPlot(params, inputs):
"""
Plots a raster from the output of batched_simulate.
Args:
params: The network parameters.
inputs: The input data for simulation (expected to be a batch, e.g., inputs[0:1]).
"""
# Get simulation traces for the provided input batch.
# If inputs has shape (1, D), then traces_batch will be (1, num_recordings, timepoints).
traces_batch = batched_simulate(params, inputs)
# Select all neuron traces for the first input in the batch.
all_neuron_traces = traces_batch[0, :, :] # Shape: (num_recordings, timepoints)
# Define simulation time parameters (from cell xPLGrxkxUVUi)
t_max = 100.0
dt = 0.025
time_axis = jnp.arange(0, t_max, dt)
# Spike detection: detect when voltage crosses a threshold from below
spike_threshold = -20.0 # Assuming a spike threshold of -20mV
spike_times_list = []
num_neurons = all_neuron_traces.shape[0]
for i in range(num_neurons): # Iterate over each neuron (recording)
neuron_trace = all_neuron_traces[i]
# Find indices where voltage crosses threshold from below
spikes = (neuron_trace[:-1] < spike_threshold) & (neuron_trace[1:] >= spike_threshold)
spike_indices = jnp.where(spikes)[0]
spike_times = time_axis[spike_indices + 1] # +1 because we are checking neuron_trace[1:]
spike_times_list.append(spike_times)
# Plotting the raster
fig, ax = plt.subplots(1, 1, figsize=(10, 5))
for i, spk_times in enumerate(spike_times_list):
ax.vlines(spk_times, i - 0.4, i + 0.4, colors='blue') # Plot vertical lines for each spike
ax.set_xlabel("Time (s)")
ax.set_ylabel("Neuron Index") # Changed label to reflect all neurons
ax.set_title("Raster Plot of All Neurons for the First Input") # Changed title
ax.set_yticks(jnp.arange(num_neurons)) # Set y-ticks to correspond to neuron indices
ax.set_ylim(-0.5, num_neurons - 0.5)
ax.set_xlim(0, t_max)
plt.grid(axis='x', linestyle='--', alpha=0.7)
plt.tight_layout()
plt.show()
# %%
# Call the rasterPlot function with the trained parameters and only the first input (as a batch)
rasterPlot(final_params, inputs[0:1])
# %%
t_max = 100.0
dt = 0.025
levels = 2
time_points = t_max // dt + 2
checkpoints = [int(np.ceil(time_points**(1/levels))) for _ in range(levels)]
def simulate(params, inputs):
currents = jx.datapoint_to_step_currents(i_delay=1.0, i_dur=1.0, i_amp=inputs / 10.0, delta_t=dt, t_max=t_max)
data_stimuli = None
data_stimuli = net.cell(0).branch(2).loc(1.0).data_stimulate(currents[0], data_stimuli=data_stimuli)
data_stimuli = net.cell(1).branch(2).loc(1.0).data_stimulate(currents[1], data_stimuli=data_stimuli)
return jx.integrate(net, params=params, data_stimuli=data_stimuli, checkpoint_lengths=checkpoints)
batched_simulate = vmap(simulate, in_axes=(None, 0))
def predict(params, inputs):
traces = simulate(params, inputs) # Shape `(batchsize, num_recordings, timepoints)`.
prediction = jnp.mean(traces[2]) # Use the average over time of the output neuron (2) as prediction.
return prediction + 72.0 # Such that the prediction is around 0.
batched_predict = vmap(predict, in_axes=(None, 0))
def predict_2(params, inputs):
"""
Calculates the mean Power Spectral Density (PSD) of the average signal of all neurons
response for frequencies in the range [0-100Hz].
"""
traces = simulate(params, inputs) # Shape `(num_recordings, timepoints)` when called by vmap
signal = jnp.mean(traces, axis=0) # Average across all neurons (axis=0)
N = signal.shape[-1] # Number of time points
fs = 1.0 / dt # Sampling frequency
# Compute one-sided FFT and corresponding frequencies for real signals
signal_fft = jnp.fft.rfft(signal)
freqs = jnp.fft.rfftfreq(N, d=dt)
# Compute Power Spectral Density (PSD)
# PSD = (1/(N*fs)) * |FFT(signal)|^2
psd = (1.0 / (N * fs)) * jnp.abs(signal_fft)**2
# Filter for frequencies in the range [0-100Hz]
mask = (freqs >= 0) & (freqs <= 100)
filtered_psd = psd[mask]
# Return the mean of the filtered PSD as a scalar prediction.
# Handle case where filtered_psd might be empty to avoid error in jnp.mean.
return jnp.mean(filtered_psd) if filtered_psd.size > 0 else 0.0
batched_predict_2 = vmap(predict_2, in_axes=(None, 0))
def loss(opt_params, inputs, labels):
params = transform.forward(opt_params)
# Use the new predict_2 function for loss calculation
predictions = batched_predict(params, inputs)
losses = jnp.abs(predictions - labels) # Mean absolute error loss.
return jnp.mean(losses) # Average across the batch.
jitted_grad = jit(value_and_grad(loss, argnums=0))
# %%
params
# %%
import jaxley.optimize.transforms as jt
# The structure passed to `jx.ParamTransform` should match the structure of `params`.
transform = jx.ParamTransform([
{"radius": jt.SigmoidTransform(0.1, 5.0)},
{"Leak_gLeak":jt.SigmoidTransform(1e-5, 1e-3)},
{"IonotropicSynapse_gS" : jt.SigmoidTransform(1e-5, 1e-2)}
])
opt_params = transform.inverse(params)
# %%
jitted_grad = jit(value_and_grad(loss, argnums=0))
value, gradient = jitted_grad(params, inputs[:4], labels[:4])
# %%
import optax
key = jax.random.PRNGKey(42)
initial_params = net.get_parameters()
# Inner optimizer (Adam) handles the gradient descent part
optimizer_inner = optax.adam(learning_rate=0.01)
# Create GSDR wrapper
optimizer = GSDR.GSDR(
inner_optimizer=optimizer_inner,
delta_distribution=jax.random.normal,
deselection_threshold=2.0,
a_init=0.4,
a_dynamic=True
)
# Initialize State
opt_state = optimizer.init(initial_params)
# %%
class Dataset:
"""A simple Dataloader which returns batches of the data.
Instead of using this simple dataloader, you can also just use one from
PyTorch or Tensorflow. You do not have to understand what is going on here
to follow this tutorial.
"""
def __init__(self, inputs: np.ndarray, labels: np.ndarray):
"""Initialize the dataloader.
Args:
inputs: Array of shape (num_samples, num_dim)
labels: Array of shape (num_samples,)
"""
assert len(inputs) == len(labels), "Inputs and labels must have same length"
self.inputs = inputs
self.labels = labels
self.num_samples = len(inputs)
self._rng_state = None
self.batch_size = 1
def shuffle(self, seed=None):
"""Shuffle the dataset in-place"""
self._rng_state = np.random.get_state()[1][0] if seed is None else seed
np.random.seed(self._rng_state)
indices = np.random.permutation(self.num_samples)
self.inputs = self.inputs[indices]
self.labels = self.labels[indices]
return self
def batch(self, batch_size):
"""Create batches of the data."""
self.batch_size = batch_size
return self
def __iter__(self):
"""Iterate over the dataset."""
self.shuffle(seed=self._rng_state)
for start in range(0, self.num_samples, self.batch_size):
end = min(start + self.batch_size, self.num_samples)
yield self.inputs[start:end], self.labels[start:end]
self._rng_state += 1
# %%
batch_size = 4
dataloader = Dataset(inputs, labels)
dataloader = dataloader.shuffle(seed=0).batch(batch_size)
# --- 5. Loop ---
key = jax.random.PRNGKey(0)
print("Starting training...")
for epoch in range(20):
key, step_key = jax.random.split(key)
epoch_loss = 0.0
for batch_ind, batch in enumerate(dataloader):
current_batch, label_batch = batch
loss_val, gradient = jitted_grad(opt_params, current_batch, label_batch)
updates, opt_state = optimizer.update(gradient, opt_state,
params=params, # Required for GSDR
value=loss_val, # Required for GSDR
key=step_key) # Required for GSDR
opt_params = optax.apply_updates(opt_params, updates)
epoch_loss += loss_val
print(f"epoch {epoch}, loss {epoch_loss}, alpha {opt_state.a}")
final_params = transform.forward(opt_params)
# %%
ntest = 32
# predictions = batched_predict(final_params, inputs[:4])
# %%
fig, ax = plt.subplots(1, 1, figsize=(3, 2))
_ = ax.scatter(labels[:ntest], predictions)
_ = ax.set_xlabel("Label")
_ = ax.set_ylabel("Prediction")
# %%
traces = batched_simulate(final_params, inputs[:4])
fig, ax = plt.subplots(1, 1, figsize=(4, 2))
_ = ax.plot(traces[:, 2, :].T)
# %% [markdown]
# # GSDR optimizer (optax standard format)
# %%
import jax
import jax.numpy as jnp
import optax
from flax.struct import dataclass
from typing import Any, Callable, NamedTuple, Optional
# State (using flax dataclass for JIT compatibility)
@dataclass
class GSDRState:
inner_state: Any # State of the inner optimizer (e.g., SGD, AdaGrad, Adam ... state)
params_opt: Any # Best parameters so far (Optimal parameters)
inner_state_opt: Any # Inner optimizer state corresponding to the optimal parameters
loss_opt: float # Optimal loss
a: float # Current self-supervision factor (alpha)
a_opt: float # Optimal self-supervision factor
def GSDR(
inner_optimizer: optax.GradientTransformation,
delta_distribution: Callable = jax.random.normal,
deselection_threshold: float = 10.0,
a_init: float = 0.5,
a_dynamic: bool = True
) -> optax.GradientTransformation:
"""
Optax-compliant implementation of the Genetic-Stochastic Delta Rule.
Args:
inner_optimizer: The gradient-based optimizer (e.g., optax.adam).
delta_distribution: Function (key, shape) -> tensor for generating noise.
deselection_threshold: Threshold factor to trigger genetic deselection.
a_init: Initial self-supervision factor (0 to 1).
a_dynamic: Whether 'a' should be stochastic/learnable.
Returns:
An optax.GradientTransformation (init_fn, update_fn).
"""
def init_fn(params):
inner_state = inner_optimizer.init(params)
return GSDRState(
inner_state=inner_state,
params_opt=params,
inner_state_opt=inner_state,
loss_opt=jnp.inf,
a=a_init,
a_opt=a_init
)
def update_fn(updates, state, params=None, value=None, key=None):
"""
Args:
updates: Gradients from loss_fn (standard Optax naming).
state: Current GSDRState.
params: Current model parameters (Required).
value: Current Loss value (Required for GSDR logic).
key: JAX PRNGKey (Required for stochastic Delta).
"""
if params is None:
raise ValueError("GSDR requires 'params' to be passed to update().")
if value is None:
raise ValueError("GSDR requires current loss 'value' to be passed to update().")
if key is None:
raise ValueError("GSDR requires a random 'key' to be passed to update().")
grads = updates
loss = value
# Split keys for delta noise and 'a' (exploration factor)
delta_key, a_key = jax.random.split(key)
# --- 1. Genetic Logic (Selection & Deselection) ---
# Optimal loss selection
is_new_opt = loss < state.loss_opt
# Update Optimal State Candidates
new_params_opt = jax.tree.map(
lambda cur, opt: jnp.where(is_new_opt, cur, opt),
params, state.params_opt
)
new_loss_opt = jnp.where(is_new_opt, loss, state.loss_opt)
new_a_opt = jnp.where(is_new_opt, state.a, state.a_opt)
# Keep the inner optimizer optimal state
new_inner_state_opt = jax.tree.map(
lambda cur, opt: jnp.where(is_new_opt, cur, opt),
state.inner_state, state.inner_state_opt
)
# Deselection (Backtrack from the Catastrophic Failure)
# If loss > threshold * best_loss, revert back to the optimal state
# Exclude the case where loss_opt is infinity (start of training)
is_deselect = (loss > (new_loss_opt * deselection_threshold)) & (new_loss_opt != jnp.inf)
# --- 2. Determine Next Step Variables ---
# If Deselecting: Revert 'a' to 'a_opt'. Else: Explore new 'a' (if a is dynamic)
if a_dynamic:
a_random = jax.random.uniform(a_key, minval=0.0, maxval=1.0)
next_a = jnp.where(is_deselect, new_a_opt, a_random)
else:
next_a = state.a # Constant
# If Deselecting: Revert inner_state to optimal. Else: Keep current.
next_inner_state = jax.tree.map(
lambda opt, cur: jnp.where(is_deselect, opt, cur),
new_inner_state_opt, state.inner_state
)
# --- 3. Calculate Updates ---
# A. Inner Optimizer Update (Gradient Descent)
# Use the *potentially reverted* inner state
inner_updates, updated_inner_state = inner_optimizer.update(grads, next_inner_state, params)
# B. Stochastic Delta Update
# Generate noise matching params structure
param_leaves, treedef = jax.tree_util.tree_flatten(params)
subkeys = jax.random.split(delta_key, len(param_leaves))
param_keys_tree = jax.tree_util.tree_unflatten(treedef, subkeys)
delta_noise = jax.tree.map(
lambda p, k: delta_distribution(k, p.shape),
params, param_keys_tree
)
# Scale noise by parameter magnitude (per paper/pseudocode)
delta = jax.tree.map(lambda n, p: n * p, delta_noise, params)
# C. Combine Updates: a * Grads + (1-a) * Delta
# Note: 'next_a' is the 'a' for THIS step.
combined_updates = jax.tree_map(
lambda d, g: next_a * d + (1 - next_a) * g,
delta, inner_updates
)
# --- 4. Handle Reset (The Revert Step) ---
# If is_deselect is True, we want the FINAL params to be params_opt.
# Optax applies: params_new = params + final_updates
# So if reset: params_new = params_opt
# Therefore: params + reset_update = params_opt
# reset_update = params_opt - params
reset_updates = jax.tree.map(
lambda opt, cur: opt - cur,
new_params_opt, params
)
# Select between Reset Update or Calculated Update
final_updates = jax.tree.map(
lambda reset, calc: jnp.where(is_deselect, reset, calc),
reset_updates, combined_updates
)
# If deselected, must NOT advance the inner optimizer state
# (use the reverted state).
# If didn't deselect, use the state returned by inner_optimizer.update
final_inner_state = jax.tree.map(
lambda reset_st, advanced_st: jnp.where(is_deselect, reset_st, advanced_st),
new_inner_state_opt, updated_inner_state
)
# Create new GSDR state
new_state = GSDRState(
inner_state=final_inner_state,
params_opt=new_params_opt,
inner_state_opt=new_inner_state_opt,
loss_opt=new_loss_opt,
a=next_a,
a_opt=new_a_opt
)
return final_updates, new_state
return optax.GradientTransformation(init_fn, update_fn)
# %%
from google.colab import drive
drive.mount('/content/drive')
from drive.MyDrive.Colab import GSDR

Jaxley_Mechanisms.ipynb at commit b4a96fa, no license · at the source

Overview

Authors: Hamed Nejat¹, Jason Sherfey², André M Bastos^1,3

ORCID iDs: Hamed Nejat, André M Bastos

Department of Psychology, Vanderbilt University, Nashville, Tennessee, United States of America
Department of Psychological and Brain Sciences, Boston University, Boston, Massachusetts, United States of America
Vanderbilt Brain Institute, Vanderbilt University, Nashville, Tennessee, United States of America

Institutions: Vanderbilt University (United States); Boston University (United States)

Journal: PloS one, volume 21, issue 8, article e0354021

Dates: received 23 March 2026; accepted 2 July 2026; published online 5 August 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.1371/journal.pone.0354021 · PMID 42555667 · PMCID PMC13440876 · OpenAlex W7172522179

Open access: gold, a free copy (OpenAlex)

Status: code verified

Categories: computational modeling (no new data) (modality), systems (subfield)

Methods: Machine learning, Spectral & time-frequency, Single-unit activity, calcium imaging

MeSH: Genetic Algorithms*, Models, Neurological*, Action Potentials, Animals, Computer Simulation, Neurons, Stochastic Processes, Visual Cortex (* major topic)

Topic: Neural dynamics and brain function (Cognitive Neuroscience, Neuroscience), according to OpenAlex

Funding: Vanderbilt University Startup Funds; NIMH NIH HHS (R00 MH116100); National Institute of Mental Health (R00MH11610); National Science Foundation (2449277)

Citations: not cited yet (Europe PMC); 132 references in the paper

Abstract

Predictive processing theories propose that the brain builds internal models of its environment by reducing the discrepancy between internally generated predictions and external sensory signals. Prior work has linked these processes to oscillatory activity in gamma (40–100 Hz) and alpha/beta (10–30 Hz) frequency ranges. Current computational approaches face a trade-off: abstract predictive-processing models can implement self-supervised computations but often omit oscillatory spiking dynamics, whereas biophysically constrained spiking models can generate neural rhythms but often require extensive manual tuning. Here, we introduce the Genetic Stochastic Delta Rule (GSDR), an evolutionary optimization framework for fitting nonlinear neural models to electrophysiological objectives. We first evaluate GSDR in simplified optimization settings, then apply it to spiking-network objectives involving firing rates, beta/gamma spectral ratios, and empirical macaque stimulus-evoked gamma dynamics from visual cortex. We show that GSDR can search constrained synaptic parameter spaces, reduce reliance on manual tuning, and reproduce spectral and circuit-level phenotypes associated with predictive routing. We also used Izhikevich simulations as a model-class robustness analysis, showing that the approach is not limited to the original Hodgkin-Huxley-style implementation. These results position GSDR as a methodological framework for automated, multi-objective exploration of oscillatory neural models, with predictive routing serving as a motivating case study rather than as a completed functional proof.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repositories

Its files are read in the Code ↔ Paper reader above, with 7 matches between paragraphs and lines of code.

HNXJ/GSDR

License: none: the authors keep all their rights

State: the link answers, verified on 27 September 2026

Evidence: files inventoried

Commit: b4a96fae5a4d00cb15889168dfc6aca76c3037ee, 6 July 2026

Languages: Python (12), Jupyter (5)

Size: 19 files, 17 scripts

Software Heritage: not archived

Found in: the supplementary material

Holds: README, 5 notebooks

Not found: license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: NumPy (12 files), JAX (11 files), Matplotlib (8 files), SciPy (5 files)

Availability: 1 check, the latest on 27 September 2026: the link answers

27 September 2026: the link answers

18 files

Biophys_SX.ipynb, Jupyter, 4,659 lines, 1 match
gsdr/__init__.py, Python, 10 lines
gsdr/analysis.py, Python, 257 lines, 1 match
gsdr/models.py, Python, 204 lines
gsdr/optimizers.py, Python, 352 lines, 1 match
gsdr/pipeline.py, Python, 170 lines
gsdr/simulation.py, Python, 82 lines
gsdr/utils.py, Python, 35 lines
kappa_synch.ipynb, Jupyter, 292 lines
misc/Biophys_SX.ipynb, Jupyter, 4,659 lines
misc/GSDR.py, Python, 15 lines
misc/GSDR_simulations.ipynb, Jupyter, 964 lines
misc/Jaxley_Mechanisms.ipynb, Jupyter, 638 lines, 4 matches
misc/jmech.py, Python, 6 lines
misc/jutils.py, Python, 12 lines
misc/main.py, Python, 1,213 lines
misc/testbench.py, Python, 35 lines
README.md, Text, 22 lines

At the source: github.com/HNXJ/GSDR

supp:PMC13440876/pone.0354021.s002.zip

License: none: the authors keep all their rights

State: the link answers, verified on 27 September 2026

Evidence: files inventoried

Languages: Python (12), Jupyter (5)

Size: 18 files, 17 scripts

Software Heritage: not checked

Found in: the supplementary material

Holds: README, 5 notebooks

Not found: license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: NumPy (12 files), JAX (11 files), Matplotlib (8 files), SciPy (5 files)

Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)

27 September 2026: the link answers (HTTP 200)

18 files

Biophys_SX.ipynb, Jupyter, 4,659 lines
gsdr/__init__.py, Python, 10 lines
gsdr/analysis.py, Python, 257 lines
gsdr/models.py, Python, 204 lines
gsdr/optimizers.py, Python, 352 lines
gsdr/pipeline.py, Python, 170 lines
gsdr/simulation.py, Python, 82 lines
gsdr/utils.py, Python, 35 lines
kappa_synch.ipynb, Jupyter, 292 lines
misc/Biophys_SX.ipynb, Jupyter, 4,659 lines
misc/GSDR.py, Python, 15 lines
misc/GSDR_simulations.ipynb, Jupyter, 964 lines
misc/Jaxley_Mechanisms.ipynb, Jupyter, 638 lines
misc/jmech.py, Python, 6 lines
misc/jutils.py, Python, 12 lines
misc/main.py, Python, 1,213 lines
misc/testbench.py, Python, 35 lines
README.md, Text, 22 lines

At the source: pmc-oa-opendata.s3.amazonaws.com/PMC13440876.1/pone.0354021.s002.zip

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

2 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
34 scripts, each with its path and the digest of its content;
7 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data Availability

https://github.com/HNXJ/GSDR https://github.com/DynaSim/DynaSim/tree/devDL.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 3 authors, 8 MeSH terms, 4 funders, 129 references.

Cite

This paper

Nejat, H., Sherfey, J., & Bastos, A. M. (2026). A genetic algorithm for self-supervised models of oscillatory neurodynamics. PloS one, 21(8), e0354021. https://doi.org/10.1371/journal.pone.0354021

BibTeX

@article{nejat2026genetic, author = {Nejat, Hamed and Sherfey, Jason and Bastos, André M}, title = {{A genetic algorithm for self-supervised models of oscillatory neurodynamics}}, journal = {PloS one}, year = {2026}, month = aug, volume = {21}, number = {8}, pages = {e0354021}, publisher = {PLOS}, issn = {1932-6203}, doi = {10.1371/journal.pone.0354021}, url = {https://doi.org/10.1371/journal.pone.0354021}, pmid = {42555667}, pmcid = {PMC13440876} }

RIS

TY - JOUR AU - Nejat, Hamed AU - Sherfey, Jason AU - Bastos, André M TI - A genetic algorithm for self-supervised models of oscillatory neurodynamics T2 - PloS one J2 - PLoS One PY - 2026 DA - 2026/08/05 VL - 21 IS - 8 SP - e0354021 SN - 1932-6203 PB - PLOS DO - 10.1371/journal.pone.0354021 UR - https://doi.org/10.1371/journal.pone.0354021 LA - en ER -

CSL-JSON

{ "id": "10.1371/journal.pone.0354021", "type": "article-journal", "title": "A genetic algorithm for self-supervised models of oscillatory neurodynamics", "container-title": "PloS one", "author": [ { "family": "Nejat", "given": "Hamed" }, { "family": "Sherfey", "given": "Jason" }, { "family": "Bastos", "given": "André M" } ], "container-title-short": "PLoS One", "volume": "21", "issue": "8", "page": "e0354021", "DOI": "10.1371/journal.pone.0354021", "PMID": "42555667", "PMCID": "PMC13440876", "ISSN": "1932-6203", "publisher": "PLOS", "URL": "https://doi.org/10.1371/journal.pone.0354021", "language": "en", "issued": { "date-parts": [ [ 2026, 8, 5 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.