Signal Propagation and Logic Gating in Networks of Integrate-and-Fire Neurons

Tim P. Vogels, L. F. AbbottJournal of Neuroscience 2005 · DOI

Abstract

Transmission of signals within the brain is essential for cognitive function, but it is not clear how neural circuits support reliable and accurate signal propagation over a sufficiently large dynamic range. Two modes of propagation have been studied: synfire chains, in which synchronous activity travels through feedforward layers of a neuronal network, and the propagation of fluctuations in firing rate across these layers. In both cases, a sufficient amount of noise, which was added to previous models from an external source, had to be included to support stable propagation. Sparse, randomly connected networks of spiking model neurons can generate chaotic patterns of activity. We investigate whether this activity, which is a more realistic noise source, is sufficient to allow for signal transmission. We find that, for rate-coded signals but not for synfire chains, such networks support robust and accurate signal reproduction through up to six layers if appropriate adjustments are made in synaptic strengths. We investigate the factors affecting transmission and show that multiple signals can propagate simultaneously along different pathways. Using this feature, we show how different types of logic gates can arise within the architecture of the random network through the strengthening of specific synapses.

Introduction

A fundamental challenge in neuroscience is understanding how signals propagate across multiple brain regions without either dissipating or escalating into runaway excitation. Historically, computational models required the injection of external "artificial" noise to maintain stability and prevent excessive synchronization. However, biological networks are inherently noisy due to their sparse, random, and balanced connectivity. This study explores whether this internal, self-generated chaos can serve as a functional resource for stable signal transmission and complex computation.

Methods

The researchers simulated a network of 10,000 leaky integrate-and-fire (LIF) neurons with a 4:1 excitatory-to-inhibitory ratio and 2% sparse connectivity. A critical component of the study was identifying the parameter space where the network remains stable and active.

Figure 1: Network stability, firing rates, and coefficient of variation as a function of excitatory and inhibitory conductance. The balanced regime (indicated by the square) provides the necessary background activity for signal propagation.

The study compared Conductance-Based (COBA) and Current-Based (CUBA) models. In COBA models, synaptic inputs modify the membrane's conductance, creating a voltage-dependent "driving force" that naturally buffers network activity. To facilitate signal propagation, specific pathways within the random architecture were identified and their synapses were strengthened by a factor of 10 to 13.

Results and Discussion

Chaotic Dynamics as a Resource

The random network generates self-sustained chaotic activity, which provides the "noise" necessary to prevent premature synchronization of rate-coded signals. This asynchronous state is vital for maintaining signal fidelity across multiple layers.

Figure 5: Spontaneous chaotic activity in the network. (a) Raster plot of 250 neurons. (b) Population activity. (c) Excitatory and inhibitory synaptic currents. (d) Membrane potential of a single neuron.

Signal Propagation and Logic Gating

The researchers found that rate-coded signals could propagate through up to six layers with high correlation (~0.6). Interestingly, the network's inherent randomness contains the structural "wiring" for various logic gates. By selectively strengthening existing synapses, the authors successfully implemented NOT and XOR gates, as well as basic memory units like flip-flops.

The COBA model proved superior to the CUBA model due to its inherent stability; as network activity increases, the effective strength of excitatory inputs decreases, preventing saturation. However, the system acts as a low-pass filter, with accuracy dropping significantly for signal frequencies above 20 Hz.

Conclusion

This work demonstrates that the "noise" of chaotic neural networks is not a hindrance but a vital component for signal transmission and computation. By leveraging the existing complexity of random networks, functional logic gates can be "discovered" and activated, suggesting that the brain's disorganized appearance may mask a highly flexible computational architecture. Future work will need to address how biological learning rules, such as STDP, might autonomously tune these pathways.

Summary updated

Abstract

Transmission of signals within the brain is essential for cognitive function, but it is not clear how neural circuits support reliable and accurate signal propagation over a sufficiently large dynamic range. Two modes of propagation have been studied: synfire chains, in which synchronous activity travels through feedforward layers of a neuronal network, and the propagation of fluctuations in firing rate across these layers. In both cases, a sufficient amount of noise, which was added to previous models from an external source, had to be included to support stable propagation. Sparse, randomly connected networks of spiking model neurons can generate chaotic patterns of activity. We investigate whether this activity, which is a more realistic noise source, is sufficient to allow for signal transmission. We find that, for rate-coded signals but not for synfire chains, such networks support robust and accurate signal reproduction through up to six layers if appropriate adjustments are made in synaptic strengths. We investigate the factors affecting transmission and show that multiple signals can propagate simultaneously along different pathways. Using this feature, we show how different types of logic gates can arise within the architecture of the random network through the strengthening of specific synapses.
BibTeX
@article{vogels2005signal,
  author = {Tim P. Vogels and L. F. Abbott},
  title = {{Signal Propagation and Logic Gating in Networks of Integrate-and-Fire Neurons}},
  journal = {Journal of Neuroscience},
  year = {2005},
  doi = {10.1523/jneurosci.3508-05.2005},
}
Figure