NEUROMORPHIC COMPUTING

We develop spiking neural network models and other learning algorithms that are compatible with neuromorphic computing platforms, with the goal of contributing to data- and energy-efficient computing and artificial intelligence.

Motivation

Neuromorphic computing offers a promising route toward energy-efficient artificial intelligence. Modern AI methods often require large computational resources, leading to high energy consumption and long computation times. In contrast, neuromorphic chips have shown substantial gains in energy-delay product for selected applications.

Despite these promising initial results, extending the success of neuromorphic computing to a wider range of applications remains a major challenge. A key limitation is the lack of learning algorithms and learning models that are broadly applicable on neuromorphic computing platforms. Our work addresses this challenge by developing machine learning and signal processing methods that draw on principles from neuromorphic technologies, such as spiking neural networks, as well as modern machine learning approaches, such as deep state-space models.

Publications:

Deep State Space Models and SNNs: 

A large part of our work explicitly utilizes a  deep state space model perspective for SNNs.  

 S. Karilanova, S. Dey, A. Özçelikkale, ``Low-Bit Data Processing Using Multiple-Output Spiking Neurons with Non-linear Reset Feedback," in IEEE Journal of Selected Topics in Signal Processing,  vol. 19, no. 6, pp. 1172-1186, Sept. 2025  (Arxiv)

 S. Karilanova, M. Fabre, E. Neftci, A. Özçelikkale, Zero-Shot Temporal Resolution Domain Adaptation for Spiking Neural Networks, Neural Networks, 2026

S. Karilanova, S. Dey, A. Özçelikkale, State-Space Model Inspired Multiple-Input Multiple-Output Spiking Neurons, Neuro Inspired Computational Elements (NICE) Conference, 2025

S. Karilanova, S. Dey, A. Özçelikkale, Delays in Spiking Neural Networks: A State Space Model Approach, 2026

S. Karilanova, S. Dey, A. Özçelikkale, Time-Varying Deep State Space Models for Sequences with Switching Dynamics, 2026

Benchmarking for neuromorphic algorithms and systems: 

Yik et. al, “The neurobench framework for benchmarking neuromorphic computing algorithms and systems”, Nature Communications, 16, vol. 1545, 2025. (Arxiv). Github Page: Neurobench Github Page. 

Applications to Tactile Systems: 

L. Chen, S.Karilanova, S.Chaki, C.Wen, L.Wang,  B. Winblad,  S.Zhang, A. Özçelikkale,  Z. Zhang “Spike-Timing Based Coding in Neuromimetic Tactile System Enables Dynamic Object Classification”, Science, 384(6696):660-665, 2024

Seminars with Recorded Videos

My PhD student Sanja's talks on our work bridging SNNs and deep SSMs:

S. Karilanova, Bridging Spiking Neural Networks and Deep State Space Models , Rise Learning Machines Seminar Series, 2026 

S. Karilanova, State-Space Model Inspired Multiple-Input Multiple-Output Spiking Neurons, Neuro Inspired Computational Elements Conference (NICE), 2025