NEUROMORPHIC COMPUTING: Rethinking Computing through Brain-Inspired Architectures
At the SMILIES group, we explore neuromorphic computing as a radical approach to overcoming the limitations of conventional architectures. By drawing inspiration from the structure and functionality of the human brain, our research aims to design energy-efficient, robust, and adaptive computing platforms tailored for edge intelligence and embedded AI.
Spiking neural networks
Hardware accelerations
Brain inspired learning
Brain inspired relability
Event based vision
Funded projects
WISE4.0 project
Web 4.0—driven by decentralisation, intelligent automation, immersive experiences, and context-aware services—will reshape most application domains including healthcare, energy, mobility, education, and media. It demands infrastructures with real-time responsiveness, trust, data integrity, and distributed intelligence.
NEUROPULS project
NEUROPULS pioneers a new era in edge computing, addressing the escalating demand for localized data processing. By harnessing neuromorphic computing principles and integrating advanced photonic security layers, NEUROPULS offers unprecedented energy efficiency, reduced latency, and enhanced security. This revolutionary project introduces novel materials for synapses/neurons and on-chip spiking sources, culminating in RISC-V compliant interfaces for seamless adaptability. With a state-of-the-art simulation platform, NEUROPULS is poised to achieve a remarkable two-fold increase in energy efficiency, setting a new standard for edge-computing systems. Join us in shaping a future where computing power meets efficiency and security head-on.
Repositories on GitHub
neurotrain
NeuroTrain is an open, extensible experimental framework that enables reproducible implementations and comparisons of SNN training algorithms under shared experimental conditions.
Elastic-Spiking-Transformers
Elastic Spiking Transformer, a novel neural network architecture that embeds runtime adaptability directly into the spiking paradigm.
EHWGesture
EHWGesture - A dataset for multimodal understanding of clinical gestures. Accepted at SAUAFG Workshop of ICCV 2025.
Spiker
Spiker is a Python-based framework for designing and generating efficient FPGA hardware accelerators for spiking neural networks, covering training, optimization, and automatic VHDL generation for edge AI deployment.
SpikingJET
SpikingJet is a full featured fault injector for Spiking Neural networks. It enables to emulate the injection of faults in an SNN model both during training and inference and assess its effect.
SpikExplorer
SpikeExplorer is a multi-objective (accuracy, latency, power) design space exploration tool able to optimize Spiking Neural Networks models to be deployed on hardware accelerators.
Publications
- A. Carpegna, A. Savino and S. Di Carlo, “Spiker: an FPGA-optimized Hardware accelerator for Spiking Neural Networks,” 2022 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), Nicosia, Cyprus, 2022, pp. 14-19. https://doi.org/10.1109/ISVLSI54635.2022.00016
- Padovano, D.; Carpegna, A.; Savino, A.; Di Carlo, S. SpikeExplorer: Hardware-Oriented Design Space Exploration for Spiking Neural Networks on FPGA. Electronics 2024, 13, 1744. https://doi.org/10.3390/electronics13091744
- B. Göğebakan, E. Magliano, A. Carpegna, A. Ruospo, A. Savino and S. D. Carlo, “SpikingJET: Enhancing Fault Injection for Fully and Convolutional Spiking Neural Networks,” 2024 IEEE 30th International Symposium on On-Line Testing and Robust System Design (IOLTS), Rennes, France, 2024, pp. 1-7. https://doi.org/10.1109/IOLTS60994.2024.10616060
Involved team
