[Paper Review] Library-based Fast Algorithm for Simulating the Hodgkin-Huxley Neuronal Networks

Zhong-qi K. Tian, Douglas Zhou|arXiv (Cornell University)|Jan 17, 2021

Neural dynamics and brain function33 references4 citations

TL;DR

This paper proposes a library-based fast algorithm for simulating Hodgkin-Huxley neuronal networks by pre-computing high-resolution data libraries of action potential dynamics, enabling large time steps (up to 0.354 ms) and achieving up to 10× speedup over standard Runge-Kutta methods while preserving key statistical properties like spike distribution and chaotic dynamics.

ABSTRACT

We present a modified library-based method for simulating the Hodgkin-Huxley (HH) neuronal networks. By pre-computing a high resolution data library during the interval of an action potential (spike), we can avoid evolving the HH equations during the spike and can use a large time step to raise efficiency. The library method can stably achieve at most 10 times of speedup compared with the regular Runge-Kutta method while capturing most statistical properties of HH neurons like the distribution of spikes which data is widely used in the statistical analysis like transfer entropy and Granger causality. The idea of library method can be easily and successfully applied to other HH-type models like the most prominent extquotedblleft regular spiking extquotedblright , extquotedblleft fast spiking extquotedblright , extquotedblleft intrinsically bursting extquotedblright{} and extquotedblleft low-threshold spike extquotedblright{} types of HH models.

Motivation & Objective

• To address the computational inefficiency of standard Runge-Kutta methods in simulating Hodgkin-Huxley (HH) neuronal networks due to stiffness during action potentials.
• To develop a fast simulation method that maintains high statistical accuracy of spike distributions and network dynamics despite using larger time steps.
• To extend the applicability of the library method to diverse HH-type models, including regular spiking, fast spiking, intrinsically bursting, and low-threshold spike neurons.
• To ensure accuracy in spike sequences under large time steps by integrating a spike-spike correction procedure to account for synaptic interaction causality.

Proposed method

• Pre-compute a high-resolution data library of membrane potential and gating variable trajectories during the action potential (3 ms) for a representative HH neuron.
• Treat the HH neuron as a leaky integrate-and-fire (I&F) model: when the membrane potential crosses threshold, suspend HH equation integration and use the library to interpolate the state after the spike.
• Use large time steps (up to 0.354 ms) during non-spiking phases, significantly reducing computational cost compared to standard Runge-Kutta with small time steps.
• Interpolate the neuron's state at the end of each time step from the pre-computed library based on the initial state at spike onset, avoiding direct integration of stiff equations.
• Apply the same library-building and lookup procedure to extended HH-type models with additional voltage-dependent currents (e.g., IB, LTS, FS types).
• Implement a spike-spike correction procedure to ensure accurate causal ordering of spikes within a time step, especially critical when using large time steps.

Experimental results

Research questions

• RQ1 Can a library-based method achieve substantial speedup in simulating Hodgkin-Huxley neuronal networks without sacrificing statistical accuracy of spike distributions?
• RQ2 To what extent can the library method preserve key dynamical properties such as spike-frequency adaptation, bursting, and chaotic behavior in HH-type models?
• RQ3 How does the library method perform across different neuronal types, including regular spiking, fast spiking, intrinsically bursting, and low-threshold spike neurons?
• RQ4 What is the maximum time step that can be used in the library method while still maintaining reliable spike sequence and statistical pattern fidelity?
• RQ5 Is the spike-spike correction procedure necessary for accurate spike sequence reconstruction when using large time steps in the library method?

Key findings

• The library method achieves a maximum of 10× speedup compared to the standard Runge-Kutta method while preserving the statistical properties of Hodgkin-Huxley neurons.
• The method maintains accurate spike distribution and firing rate statistics, which are essential for transfer entropy and Granger causality analyses requiring long simulation runs (~20 minutes).
• The library method successfully captures chaotic dynamics, as confirmed by the largest Lyapunov exponent, indicating retention of complex temporal behavior.
• The method is robust across diverse HH-type models, including regular spiking, fast spiking, intrinsically bursting, and low-threshold spike neurons, with consistent speedup and accuracy.
• The spike-spike correction procedure is essential for accurate spike sequence reconstruction when using large time steps, ensuring correct synaptic interaction causality.
• Even with extra error from library interpolation, the method retains most statistical and dynamical properties of the original HH model, making it suitable for long-term statistical neuroscience applications.