Fit Grid Cells population

import math
import os
from typing import Optional

import jax
import matplotlib.pyplot as plt
import nemos as nmo
import numpy as np
import pynapple as nap
import nemos as nmo
from scipy.ndimage import gaussian_filter

import workshop_utils

jax.config.update("jax_enable_x64", True)

DATA STREAMING

Stream the data

io = workshop_utils.data.download_dandi_data("000582", "sub-11265/sub-11265_ses-07020602_behavior+ecephys.nwb",
)

PYNAPPLE

Load the data with pynapple

data = nap.NWBFile(io.read())

Print the data

print(data)

extract the spike times and the position of the animal frome the data object

spikes = data["units"]  # Get spike timings
position = data["SpatialSeriesLED1"] # Get the tracked orientation of the animal

compute the head-direction of the animal from SpatialSeriesLED1 and SpatialSeriesLED1

diff = data['SpatialSeriesLED1'].values-data['SpatialSeriesLED2'].values
head_dir = (np.arctan2(*diff.T) + (2*np.pi))%(2*np.pi)
head_dir = nap.Tsd(data['SpatialSeriesLED1'].index, head_dir).dropna()

compute the head-direction and position tuning curves

hd_tuning = nap.compute_1d_tuning_curves(
    group=spikes, 
    feature=head_dir,
    nb_bins=61, 
    minmax=(0, 2 * np.pi)
    )

pos_tuning, binsxy = nap.compute_2d_tuning_curves(
    group=spikes, 
    features=position, 
    nb_bins=12)

plot the tuning curves for each neurons

fig = plt.figure(figsize = (12, 4))
gs = plt.GridSpec(2, len(spikes))
for i in range(len(spikes)):
    ax = plt.subplot(gs[0,i], projection='polar')
    ax.plot(hd_tuning.loc[:,i])

    ax = plt.subplot(gs[1,i])
    ax.imshow(gaussian_filter(pos_tuning[i], sigma=1))
plt.tight_layout()

NEMOS

bin spike trains in 10 ms bin size

interpolate the position to the timestamps of counts using interpolate function of pynapple

define a basis in 2D using nemos RaisedCosineBasisLinear

evaluate the basis on a 100x100 grid using evaluate_on_grid

plot the evaluated basis

rescale the position between 0 and 1 to match the basis functions

evaluate the basis for each position of the animal

instantiate a GLM model with Ridge regularization.
set regularizer_strength=1.0

fit the model only to neuron 7 for faster computation

predict the rate and compute a tuning curves using compute_2d_tuning_curves_continuous from pynapple

compare the tuning curves

find the best regularizer_strength using sklearn.model_selection.GriSearchCV

instantiate the best model from scikit-learn

predict the rate of the best model and compute a 2d tuning curves

compare the 2d tuning curves

Total running time of the script: ( 0 minutes 0.000 seconds)

Download Python source code: 03_grid_cells_users.py

Download Jupyter notebook: 03_grid_cells_users.ipynb

Gallery generated by mkdocs-gallery