Why to use NumpyBrain

NumpyBrain is a microkernel framework for SNN (spiking neural network) simulation purely based on native python. It only relies on NumPy. However, if you want to get faster performance,you can additionally install Numba. With Numba, the speed of C or FORTRAN can be obtained in the simulation.

NumpyBrain wants to provide a highly flexible and efficient SNN simulation framework for Python users. It endows the users with the fully data/logic flow control. The core of the framework is a micro-kernel, and it’s easy to understand (see How NumpyBrain works). Based on the kernel, the extension of the new models or the customization of the data/logic flows are very simple for users. Ample examples (such as LIF neuron, HH neuron, or AMPA synapse, GABA synapse and GapJunction) are also provided. Besides the consideration of flexibility, for accelerating the running speed of NumPy codes, Numba is used. For most of the times, models running on Numba backend is very fast (see examples/benchmark).

More details about NumpyBrain please see our document.

Installation

Install NumpyBrain using pip:

$> pip install git+https://github.com/PKU-NIP-Lab/NumpyBrain

Install from source code:

$> python setup.py install

The following packages need to be installed to use NumpyBrain:

• Python >= 3.5

• NumPy >= 1.13

• Sympy >= 1.2

• Matplotlib >= 2.0

• autopep8

Packages recommended to install:

• Numba >= 0.40.0

• JAX >= 0.1.0

Define a Hodgkin–Huxley neuron model

import npbrain.numpy as np
import npbrain as nb

def HH(noise=0., E_Na=50., g_Na=120., E_K=-77., g_K=36.,
       E_Leak=-54.387, g_Leak=0.03, C=1.0, Vth=20.):

    ST = nb.types.NeuState(
        {'V': -65., 'm': 0., 'h': 0., 'n': 0., 'sp': 0., 'inp': 0.},
        help='Hodgkin–Huxley neuron state.\n'
             '"V" denotes membrane potential.\n'
             '"n" denotes potassium channel activation probability.\n'
             '"m" denotes sodium channel activation probability.\n'
             '"h" denotes sodium channel inactivation probability.\n'
             '"sp" denotes spiking state.\n'
             '"inp" denotes synaptic input.\n'
    )

    @nb.integrate
    def int_m(m, t, V):
        alpha = 0.1 * (V + 40) / (1 - np.exp(-(V + 40) / 10))
        beta = 4.0 * np.exp(-(V + 65) / 18)
        return alpha * (1 - m) - beta * m

    @nb.integrate
    def int_h(h, t, V):
        alpha = 0.07 * np.exp(-(V + 65) / 20.)
        beta = 1 / (1 + np.exp(-(V + 35) / 10))
        return alpha * (1 - h) - beta * h

    @nb.integrate
    def int_n(n, t, V):
        alpha = 0.01 * (V + 55) / (1 - np.exp(-(V + 55) / 10))
        beta = 0.125 * np.exp(-(V + 65) / 80)
        return alpha * (1 - n) - beta * n

    @nb.integrate(noise=noise / C)
    def int_V(V, t, m, h, n, Isyn):
        INa = g_Na * m ** 3 * h * (V - E_Na)
        IK = g_K * n ** 4 * (V - E_K)
        IL = g_Leak * (V - E_Leak)
        dvdt = (- INa - IK - IL + Isyn) / C
        return dvdt

    def update(ST, _t_):
        m = np.clip(int_m(ST['m'], _t_, ST['V']), 0., 1.)
        h = np.clip(int_h(ST['h'], _t_, ST['V']), 0., 1.)
        n = np.clip(int_n(ST['n'], _t_, ST['V']), 0., 1.)
        V = int_V(ST['V'], _t_, m, h, n, ST['inp'])
        sp = np.logical_and(ST['V'] < Vth, V >= Vth)
        ST['sp'] = sp
        ST['V'] = V
        ST['m'] = m
        ST['h'] = h
        ST['n'] = n
        ST['inp'] = 0.

    return nb.NeuType(requires={"ST": ST}, steps=update, vector_based=True)

Define an AMPA synapse model

def AMPA(g_max=0.10, E=0., tau_decay=2.0):

    requires = dict(
        ST=nb.types.SynState(['s'], help='AMPA synapse state.'),
        pre=nb.types.NeuState(['sp'], help='Pre-synaptic state must have "sp" item.'),
        post=nb.types.NeuState(['V', 'inp'], help='Post-synaptic neuron must have "V" and "inp" items.')
    )

    @nb.integrate(method='euler')
    def ints(s, t):
        return - s / tau_decay

    def update(ST, _t_, pre):
        s = ints(ST['s'], _t_)
        s += pre['sp']
        ST['s'] = s

    @nb.delayed
    def output(ST, post):
        post_val = - g_max * ST['s'] * (post['V'] - E)
        post['inp'] += post_val

    return nb.SynType(requires=requires, steps=(update, output), vector_based=False)