Spiking Neural Networks

Spiking Neural Networks

Neuromorphic perception stacks that process biosignals, tactile data, and event streams through sparse spikes for ultra-low-power intelligence.

• Biologically grounded neuron models tuned with clinical priors
• Tooling that converts ANN checkpoints into deployable SNN graphs
• Edge deployments on Loihi, FPGA accelerators, and custom mixed-signal chips

Research Area Overview

Spiking Neural Networks (SNNs) communicate through 1-bit spikes, mirroring the event-driven nature of biological neurons. This spike-based computation compresses models, reduces energy consumption, and stays close to neuroscience insights, making SNNs ideal for edge devices and neuromorphic chips.

What We Explore

• Bio-plausible neuron and synapse models rooted in computational neuroscience
• Conversion and training techniques that maintain accuracy while leveraging sparse spike events
• Interfaces between SNNs and neuromorphic sensors (event cameras, tactile skins, bio-signal front-ends)

Lab Focus

Our lab analyzes biosignals such as ECG via SNNs and refines network topologies using prior neuroscience studies. The ultimate goal is to build neuromorphic systems that learn, adapt, and explain their decisions without heavy compute.