Modeling of Self-sustained Neuron Population without External Stimulus

A research team led by İhsan Ertuğrul Karakaş has published groundbreaking findings on arXiv demonstrating that biologically-inspired neural networks can sustain autonomous activity without continuous external stimulation. Their paper, 'Modeling of Self-sustained Neuron Population without External Stimulus,' details a simulated network of 200 Hodgkin-Huxley neurons (160 excitatory, 40 inhibitory) incorporating sophisticated biological mechanisms including spike-timing-dependent plasticity (STDP), probabilistic vesicle release, and receptor variability. After receiving only a brief 200ms initialization stimulus to 30 excitatory neurons, the network operated completely autonomously for extended periods.

In their primary 1800-second (30-minute) simulation, the network maintained sparse, irregular firing patterns with 67% of neurons firing below 1 Hz and a population mean firing rate of 1.13 ± 1.34 Hz. The researchers observed that participation increased over longer observation windows, and population-mean Fano factors remained between 1-2, indicating irregular spike timing characteristic of biological neural systems. Remarkably, the network showed spontaneous qualitative reorganizations in collective firing patterns over time, suggesting emergent dynamics rather than simple repetitive loops.

The findings challenge conventional assumptions about neural network stability and suggest that recurrent networks with plastic and stochastic synapses can maintain long-duration autonomous activity in sparse firing regimes. This work bridges computational neuroscience and artificial intelligence by demonstrating how biological principles like STDP and stochasticity can enable self-sustaining computation, potentially informing more energy-efficient and biologically-plausible AI architectures that don't require constant external input.