Great question! In artificial neural networks, neurons rely on activation functions to introduce the non-linearity needed for learning. A classic example is the sigmoid activation function, and here's a straightforward Python implementation:

import numpy as np

def sigmoid(x):
    return 1 / (1 + np.exp(-x))

Now, if you're curious about implementing a biological activation function like the Hodgkin-Huxley (HH) model— which accurately models the electrical behavior of excitable cells such as neurons—here's a breakdown:

Mathematical Form of the Hodgkin-Huxley Model #

The core differential equation of the HH model is:
$$C_m \frac{dV_m}{dt} = I - g_{Na}m^3h(V_m - E_{Na}) - g_Kn^4(V_m - E_K) - g_L(V_m - E_L)$$

Where the key parameters are defined as:

• $C_m$: Membrane capacitance (measures the cell membrane's ability to store electrical charge)
• $V_m$: Membrane potential (the electrical potential difference across the cell membrane)
• $I$: External applied current
• $g_{Na}, g_K, g_L$: Conductances of sodium, potassium, and leak ion channels respectively
• $E_{Na}, E_K, E_L$: Reverse potentials for each corresponding channel type
• $m, h, n$: Gating variables that describe the opening/closing dynamics of ion channels over time

As described in the classic Hodgkin-Huxley framework, each component of an excitable cell is treated as an electrical element. The lipid bilayer acts as a capacitor ($C_m$), ion channels function as variable resistors (with conductances dependent on membrane potential and time), and ion pumps maintain the concentration gradients that drive ion flow across the membrane.

To fully implement the HH model, you also need to account for the time-dependent gating variables, which have their own differential equations:
$$\frac{dm}{dt} = \alpha_m(V_m)(1 - m) - \beta_m(V_m)m$$
$$\frac{dh}{dt} = \alpha_h(V_m)(1 - h) - \beta_h(V_m)h$$
$$\frac{dn}{dt} = \alpha_n(V_m)(1 - n) - \beta_n(V_m)n$$

Here, $\alpha$ and $\beta$ are voltage-dependent rate constants that determine how quickly the gating variables open or close. In practice, you'd use an ODE solver like scipy.integrate.solve_ivp in Python to numerically solve this system of equations over a time interval and simulate the membrane potential dynamics.

内容的提问来源于stack exchange，提问作者Larry