Spiking neuron reservoirs improve frequency detection for better AI accuracy
Beyond Low-Pass Dynamics: Frequency-Selective Spiking Reservoirs with Resonant Neurons
Artificial Intelligence
Summary
Modern spiking neural networks often use simple neurons that blur fast changes in signals, making it hard to tell apart different rhythms in data. The authors replaced these simple neurons with special ones that act like tiny oscillators tuned to different frequencies, helping the network distinguish patterns that happen at specific rhythms. This change makes the system better at recognizing patterns in things like sound, images, and sensor data while still using very little energy. Their new design performs better than earlier models on many tests without draining extra power.
What this means in practice
- •For embedded system engineers: Build low-energy devices that better recognize time-based signals using frequency-selective spiking neuron reservoirs.
- •For audio signal processing teams: Improve classification of audio spike trains by deploying spiking reservoirs that respond selectively to input frequencies.
Authors
Aurora Micheli, Claudio Gallicchio, Nergis Tomen
Abstract
Spiking reservoir computing combines the training efficiency of fixed recurrent reservoirs with the energy efficiency of event-driven inference, where computation is triggered by sparse spikes rather than executed densely at every timestep. However, existing spiking reservoirs almost exclusively rely on Leaky Integrate-and-Fire (LIF) neurons, whose low-pass dynamics limit their ability to distinguish information encoded at different temporal frequencies. We introduce HRF-Res, a liquid state machine in which LIF neurons are replaced by a heterogeneous population of Harmonic Resonate-and-Fire (HRF) neurons. These neurons exhibit intrinsic oscillatory dynamics and selectively respond to input frequencies, enabling the reservoir to capture frequency-specific temporal patterns. The result is more discriminative representations and higher classification accuracy than LIF-based reservoirs, achieved without sacrificing the energy efficiency of spiking computation, with up to 19x lower energy consumption than an equivalent non-spiking oscillatory reservoir. Evaluated on ten benchmarks spanning time series, sequential images, audio spike trains, and neuromorphic vision, HRF-Res achieves state-of-the-art accuracy among spiking reservoir methods. These results demonstrate that introducing frequency-selective dynamics into spiking reservoirs improves representational quality and classification performance while preserving their inherent energy efficiency.