<ul><li>The Spiking Neural Network (SNN) has gained attention for its energy-efficient and biological plausible processing.</li><li>Training SNNs involves using surrogate gradients to approximate the non-differentiable spike function near the firing threshold.</li><li>A challenge called the 'dilemma of gamma' arises due to the surrogate gradient support width affecting overactivation in neurons.</li><li>To tackle this challenge, a temporal Inhibitory Leaky Integrate-and-Fire (ILIF) neuron model is proposed, which reduces overactivation, enhances energy efficiency, and stabilizes training.</li></ul>

ILIF: Temporal Inhibitory Leaky Integrate-and-Fire Neuron for Overactivation in Spiking Neural Networks

Discover more