Microcircuit firing rates

Here we calculate the firing rates of the Potjans and Diesmann [2012] microcircuit model.

import nnmt
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches

# import svgutils for inserting sketch if available
try:
    import os
    import svgutils.transform as sg
    insert_sketch = True
except ImportError:
    insert_sketch = False

# use matplotlib style file
plt.style.use('frontiers.mplstyle')

# decide output format (either pdf or eps)
ext = 'pdf'

First we create a network model of the microcircuit, passing the parameter yaml file.

microcircuit = nnmt.models.Microcircuit(
    '../../tests/fixtures/integration/config/Bos2016_network_params.yaml')

Then we simply calculate the firing rates for exponentially shape post synaptic currents, by calling the respective function and passing the microcircuit. Here we chose to use the ‘taylor’ method for calculating the firing rates.

firing_rates = nnmt.lif.exp.firing_rates(microcircuit, method='shift')

print(f'Mean rates: {firing_rates}')

Then we compare the rates to the publicated data from Bos et al. [2016]. We load the simulated rates using the data stored as integration test fixtures. Note that the original data use rates in 1/ms, which we need to convert to Hz.

fix_path = '../../tests/fixtures/integration/data/'
result = nnmt.input_output.load_h5(
    fix_path + 'Bos2016_publicated_and_converted_data.h5')
simulated_rates = result['fig_microcircuit']['rates_sim'] * 1000

print(f'Mean simulated rates: {simulated_rates}')

Finally, we plot the rates together in one plot.

width = 0.03937007874 * 85
height = width * 1.3

fig = plt.figure(figsize=(width, height))

sim_colors = ['#4c72b0', '#c44e52']
thy_color = '#ff8f2fff'

ax0 = plt.subplot2grid((5, 1), (0, 0), rowspan=3, colspan=1)
ax0.set_axis_off()
ax1 = plt.subplot2grid((5, 1), (3, 0), rowspan=2, colspan=1)

# add labels to panels
axs = [ax0, ax1]
label = ['A', 'B']
y_pos = [1.25, 1.1]
for n, ax in enumerate(axs):
    ax.text(-0.1, y_pos[n], label[n], transform=ax.transAxes,
            size=11, weight='bold')


bars = ax1.bar(np.arange(8), simulated_rates,
               align='center', color=sim_colors[1])

for i in [0, 2, 4, 6]:
    bars[i].set_color(sim_colors[0])

nnmt_handle = ax1.scatter(np.arange(8), firing_rates, marker='X',
                          color=thy_color, s=50, zorder=10)
ax1.set_xticks(np.arange(8))
ax1.set_xticklabels(['2/3E', '2/3I', '4E', '4I', '5E', '5I', '6E', '6I'])
ax1.set_yticks([1, 3, 5, 7])
ax1.set_ylabel(r'rate $\nu\,(1/s)$')

plt.legend([bars[0], nnmt_handle, bars[1]],
           [None, 'theory', 'simulation'],
           loc='upper left', fontsize=9, ncol=2,
           columnspacing=-2.8, handletextpad=0.2)

# insert sketch using svgutil, try saving as pdf using inkscape
if insert_sketch:
    sketch_fn = 'microcircuit_sketch.svg'
    plot_fn = 'microcircuit_rates'
    svg_mpl = sg.from_mpl(fig, savefig_kw=dict(transparent=False))
    w_svg, h_svg = svg_mpl.get_size()
    svg_mpl.set_size((w_svg+'pt', h_svg+'pt'))
    svg_sketch = sg.fromfile(sketch_fn).getroot()
    svg_sketch.moveto(x=30, y=10, scale_x=0.61, scale_y=0.61)
    svg_mpl.append(svg_sketch)
    svg_mpl.save(f'{plot_fn}.svg')
    if ext == 'pdf':
        os_return = os.system(f'inkscape --export-pdf={plot_fn}.pdf {plot_fn}.svg')
    elif ext == 'eps':
        os_return = os.system(f'inkscape {plot_fn}.svg -E {plot_fn}.eps --export-ps-level=3')
    else:
        print(f'Unknown output file extention `{ext}`. '
              'Please choose either `pdf` or `eps`.')
        os_return = -1
    if os_return == 0:
        os.remove(f'{plot_fn}.svg')
    else:
        print('Conversion to pdf or eps using inkscape failed, keeping svg...')

ax0.annotate('(sketch)', xy=(0.35, 0.6))
plt.show()