Welcome

In today's data-driven world, rapid, reliable, and accurate decision-making through the processing of vast amounts of multi-modal sensory data, all under strict low-power budgets, is essential for many applications, including edge computing, medical devices, autonomous robotics, and remote monitoring systems. However, as device scaling continues, hardware errors increase, prompting the electronic systems community to explore novel approaches for hardware reliability. The growing demand for high reliability and fault tolerance, coupled with the need for small size, lightweight, and low-power consumption (SWaP), elevates dependability and low-SWaP as critical design objectives.

Moreover, sustainability has become a critical R&D challenge in electronic systems, as mitigating faults and ensuring a system's continued correct operation extends its functional lifetime. Addressing faults early effectively contributes to the long-term sustainability of these systems, reducing the frequency of replacements and minimising electronic waste.

Achieving these design goals requires a transformative approach. The human brain provides a novel perspective with its unique ability to robustly perform real-time information processing and integration, at multiple scales and levels of abstraction. Particularly, its plasticity and adaptation capability can handle decisions in dynamic and uncertain environments. Despite advances in computational neuroscience, our understanding of the brain remains incomplete, let alone realising brain-like performance in electronic systems. Combining data and models from different contexts and different spatial and temporal scales remains an open challenge.

While the successes of deep learning have been rapidly growing, they are high in power consumption; still restricted to the specific fields of image recognition, language processing, and game-playing; and are computationally dissimilar to brain-like information processing.

Event-based computational models such as spiking neural networks (SNN) are closer to biology, and we have witnessed progress in efficiently mimicking SNN hardware to the point where scalable building blocks of synapses and neurons are now available.

Neuromorphic engineering combines neuroscience and neural network research, bridging the gap between fundamental research and practical engineering solutions by offering efficient and fast computing devices (e.g., Loihi, TrueNorth, SpiNNaker, BrainScaleS). Despite these significant developments, an absence of scalable design methods and design tools, the field is still early in its journey in terms of meeting real-world application requirements.

The goal of this Special Session is to present and discuss research from world-leading experts addressing state-of-the-art design techniques and methodologies to achieve reliability, efficiency, and scalability in, and with, neuromorphic hardware. Focussing on edge, embedded, and autonomous systems where these performance metrics are paramount, the session includes current approaches to bio-inspired fault tolerance in SoCs, design automation, and applications in robotics and edge computing. As such, this Special Session is of strong relevance to WCCI’s tracks IJCNN and CEC. Addressing these challenges will contribute to the long-term sustainability of electronic systems by ensuring they operate efficiently and reliably over extended periods, thus extending their useful lifetime.

Session Papers Submission

Please make sure to select IJCNN SS40 Design Challenges, Methods and Applications in Sustainable Neuromorphic Hardware when subitting your paper.

IJCNN SS40 Submit Paper

Organisation Team