FiTS: Interpretable Spiking Neurons via Frequency Selectivity and Temporal Shaping

2026-05-13

AI Summary

Traditional spiking neurons lack explicit modeling of temporal specialization mechanisms at the individual neuron level, making it challenging to balance performance and interpretability. This work proposes the FiTS spiking neuron model, which, for the first time, decouples temporal computation within a single neuron into two distinct modules: frequency selectivity (FS) and temporal shaping (TS), governing frequency preference and the temporal structure of membrane potential accumulation, respectively. Leveraging FS parameterization and group delay modulation, FiTS enables feedforward spiking neural networks without recurrent connections or network-level delays. On auditory benchmark tasks, FiTS significantly outperforms standard leaky integrate-and-fire (LIF) models and matches state-of-the-art temporal spiking neural network approaches, while offering highly interpretable, neuron-level frequency and temporal response characteristics.

Abstract

Spiking Neural Networks (SNNs) are a promising framework for event-driven temporal processing. Prior work has improved temporal modeling through richer neuron dynamics and network-level mechanisms such as recurrence and delays, but it remains unclear how individual spiking neurons should specialize within a network. In this work, we introduce FiTS, a spiking neuron that factorizes temporal computation within each neuron into Frequency Selectivity (FS) and Temporal Shaping (TS). The FS module parameterizes each neuron's target frequency as the maximizer of its subthreshold magnitude response, while the TS module reshapes when frequency components contribute to membrane voltage accumulation through group-delay modulation. On auditory benchmarks where frequency selectivity and timing are central to the input structure, FiTS consistently improves over a plain Leaky Integrate-and-Fire (LIF) baseline in simple feedforward SNNs without recurrence or network-level delays, while remaining competitive with strong temporal SNN baselines. Beyond accuracy, the learned target frequencies and group-delay shifts provide interpretable neuron-level summaries of the frequency and timing organization learned within the network.

Problem

• Spiking Neural Networks
• Frequency Selectivity
• Temporal Processing
• Neuron Specialization
• Interpretability

Innovation

• Frequency Selectivity
• Temporal Shaping
• Spiking Neural Networks
• Interpretability
• Group-delay Modulation