Neuromorphic Computing: Current Progress and the Future of Brain-Inspired Computing

NEUROMORPHIC COMPUTING AND PRACTICAL MEDIUM DEPENDENCE

Neuromorphic Computing

- The neuromorphic computing is a process that mimics the processing of information at biological synapses. The physical mechanism of neuromorphic computing is based on the ion transport at device interface to achieve efficient transfer and computation of information at the same location [11,12], which is different from logic circuit in digital systems [13,14]. Besides, neuromorphic computing is characterized by distributed memory and computational elements [15]. The neuromorphic computing can realize high-efficiency massive parallel computing that can conquer the limitation of Von Neumann bottleneck. Neuromorphic computing is a paradigm of designing hardware and algorithms inspired by the brain’s architecture and principles, promising major gains in energy efficiency and new computing capabilities. This review provides a comprehensive overview of developments in neuromorphic computing from 2019 through 2024. We survey hardware advances – including digital neuromorphic chips (e.g. Intel Loihi, IBM TrueNorth, and SpiNNaker), emerging device technologies like memristors, spintronic circuits, photonic processors, and two-dimensional (2D) material-based devices – that enable brain-like computation with vastly lower power than conventional electronics. We also summarize algorithmic advances in spiking neural networks (SNNs), covering progress in temporal coding strategies, the introduction of surrogate gradient methods for training SNNs like deep networks, and biologically plausible learning rules such as e-prop for online learning in spiking systems. Furthermore, we discuss key opportunities and gaps: the potential of neuromorphic systems to approach aspects of human cognition or artificial general intelligence (AGI), applications in medicine (like brain – machine interfaces and neural prosthetics) and science, the trade-offs between power efficiency and computational precision, and challenges in integrating neuromorphic accelerators into existing computing ecosystems. We conclude by highlighting how co-development of hardware and algorithms is critical to fulfill the promise of neuromorphic computing, and by outlining open research directions on the path toward more brain-like, efficient computing architectures.

Spiking Neural Networks: A Computational Paradigm for Neuromorphic Computing

It is aimed at proving that SNNs have potential in such areas as computer vision, robotics, and speech recognition, and their role in overcoming the barrier between artificial and biological neural systems is proved.

Neuromorphic RISC-V Systems for Bio-inspiredComputing Applications

Neuromorphic computing is a paradigm based on the computational mechanisms of the human brain and hasreceived considerable attention as a real-time technique with low energy requirements. Present systems, however, arelimited in their ability to scale traditional processors to a neuromorphic architecture, leading to issues with latency,power consumption, and smooth data flow. To address these problems, this paper proposes the ACORISC-VbSNNframework, comprising a modular RISC-V architecture, Spiking Neural Networks (SNNs), and Ant ColonyOptimization (ACO). The system uses a shared-memory architecture to maximize communication between traditionaland neuromorphic processors, ensuring data is managed effectively. The postulated framework processes the sensorydata by pre-processing and encoding them using rate coding, and dynamically optimizing memory access. SNNs arealso used to process spike trains in real-time, whereas ACO is used to determine the best data paths to minimizebottlenecks. Experimental analysis shows that the system performs better, with ultra-low power consumption of 0.0095mW, very low latency of 0.000544 seconds, and 99.2 percent accuracy. These findings indicate that the ACORISCVbSNN model has the potential to advance the field of bio-inspired computing, providing a highly accurate, energyefficient, and low-latency system for real-world use.

Neuromorphic architectures as numerical solvers for computational neuroscience

Neuromorphic computing is closely associated with spiking neuronal networks. However, an alternative class of so-called"rate-based"models arising from computational neuroscience and machine learning forgoes spiking interactions and instead relies on continuous coupling between neurons. Existing neuromorphic implementations designed around spike-based interactions are not well-suited for emulating such models. Here view the distributed simulation of these models as message-passing algorithms on parallel hardware. Leveraging prior art in numerical algorithms and distributed simulation, we outline steps that enable the design of efficient digital neuromorphic accelerators for non-spiking neuronal models. In particular, we show that multi-bit packets, rather than spikes, are the most efficient communication strategy in packet-switched networks and that compared to basic numerical integration methods, higher-order differential equation solvers decrease both computation and communication costs while achieving lower numerical error, but that these benefits are ultimately limited by arithmetic precision. Using our proposed design principles, we convert an existing neuromorphic architecture into a distributed numerical solver - a spikeless neuromorphic system - for continuously-coupled neuronal models. We thereby demonstrate that our theoretical considerations indeed translate into practical advantages, namely reduced energy consumption and delay.

( Invited ) Cryogenic Electrochemical Ion Transport and Switching Dynamics In Solid-State Electronic Synapses

The rapid advancement of artificial intelligence (AI) has driven significant progress in data processing, pattern recognition, and autonomous decision-making across various domains, including healthcare, smart manufacturing, and scientific computing. These advances increasingly rely on large-scale artificial neural networks, whose expanding model sizes demand substantial computational and memory resources. However, conventional von Neumann architectures, which physically separate processing and memory, suffer from intrinsic limitations including severe data-transfer bottlenecks. As a result, global data-center electricity consumption is projected to reach ~945 TWh by 2030, underscoring the urgent need for energy-efficient computing paradigms. Although software-level approaches such as model compression, quantization, and pruning can partially reduce computational cost, they cannot fundamentally overcome the hardware inefficiencies of von Neumann systems. This challenge has motivated the development of neuromorphic computing based on non–von Neumann architectures, which co-locate memory and computation to emulate brain-inspired information processing. In particular, cryogenic neuromorphic computing has emerged as a promising direction, especially when integrated with quantum computing (QC). Cryogenic operation offers reduced thermal noise and power dissipation, enabling high performance and ultra-low-power computation. However, such systems require electronic synapses that can be placed close to quantum processors, reliably store quantum-generated information, and operate efficiently at cryogenic temperatures. Spiking neural networks (SNNs) offer a biologically inspired framework well-suited for this purpose. In SNNs, neurons integrate inputs and emit spikes upon reaching a threshold, while synapses adapt their weights through spike-timing-dependent plasticity (STDP). This event-driven, time-domain encoding enables highly energy-efficient computation and naturally unifies memory and processing. In contrast, conventional memories such as static random access memory (SRAM), dynamic random access memory (DRAM), and NOR/NAND Flash are optimized for digital storage rather than analog synaptic weight modulation. Their discrete states, high power consumption, limited endurance, and strict logic-memory separation significantly hinder their suitability for neuromorphic systems, particularly under cryogenic operation. Emerging memory technologies capable of continuous conductance modulation and low-energy operation are therefore essential for electronic synapses. Among them, resistive RAM (RRAM), electrochemical RAM (ECRAM), and conductive-bridge RAM (CBRAM) have garnered significant attention due to their simple structures, scalability, and low switching energy. ECRAM relies on ionic drift mechanisms analogous to biological synapses, while CBRAM modulates resistance via the formation and rupture of metallic filaments, enabling analog conductance control. In this work, we investigate cryogenic electronic synapses based on LiCoO 2 (LCO) and Ag/WTe 2 CBRAM structures over a wide temperature range from cryogenic to room temperature. For the LCO synapse, an Al-rich LiCoO 2 (LACO) Li-ion nano-reservoir stabilizes Li ions, reduces interfacial electric fields, and suppresses leakage current. At cryogenic temperatures, suppressed ion and electron diffusion enhances field-driven processes, resulting in significantly improved synaptic performance, including an expanded memory window (~1.6 to ~423), improved linearity, and enhanced STDP. We further demonstrate an Ag/WTe 2 CBRAM synaptic device featuring forming-free switching, self-compliance, and an ultra-low operating voltage of ~0.2 V. The WTe 2 matrix enables stable Ag filament formation, supporting key synaptic functions such as short-term plasticity, long-term potentiation, and STDP. I-V measurements from 4 K to 300 K reveal suppressed memory behavior below 10 K due to quenched thermally activated Ag + migration. In contrast, memory windows and resistance oscillations emerge at 20–30 K, which are likely attributed to electron hopping during the early stages of filament nucleation. These results decouple thermal and field-driven effects, providing critical guidance for designing reliable cryogenic electronic synapses for future neuromorphic-quantum computing systems.