Temporal dendritic heterogeneity incorporated with spiking neural networks for learning multi-timescale dynamics

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Jan. 4, 2024

It is widely believed the brain-inspired spiking neural networks have capability of processing temporal information owing to their dynamic attributes. However, how understand what kind mechanisms contributing learning ability and exploit rich properties satisfactorily solve complex computing tasks in practice still remains be explored. In this article, we identify importance capturing multi-timescale components, based on which a multi-compartment model with dendritic heterogeneity, proposed. The enables dynamics by automatically heterogeneous timing factors different branches. Two breakthroughs are made through extensive experiments: working mechanism proposed revealed via an elaborated XOR problem analyze feature integration at levels; comprehensive performance benefits over ordinary achieved several benchmarks for speech recognition, visual electroencephalogram signal robot place shows best-reported accuracy compactness, promising robustness generalization, high execution efficiency neuromorphic hardware. This work moves significant step toward real-world applications appropriately exploiting biological observations.

Language: Английский

---

Neuromorphic hardware for somatosensory neuroprostheses

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: Jan. 16, 2024

Abstract In individuals with sensory-motor impairments, missing limb functions can be restored using neuroprosthetic devices that directly interface the nervous system. However, restoring natural tactile experience through electrical neural stimulation requires complex encoding strategies. Indeed, they are presently limited in effectively conveying or sensations by bandwidth constraints. Neuromorphic technology, which mimics behavior of neurons and synapses, holds promise for replicating touch, potentially informing neurostimulation design. this perspective, we propose incorporating neuromorphic technologies into neuroprostheses could an effective approach developing more human-machine interfaces, leading to advancements device performance, acceptability, embeddability. We also highlight ongoing challenges required actions facilitate future integration these advanced technologies.

Language: Английский

---

DYNAP-SE2: a scalable multi-core dynamic neuromorphic asynchronous spiking neural network processor

Neuromorphic Computing and Engineering, Journal Year: 2024, Volume and Issue: 4(1), P. 014003 - 014003

Published: Jan. 9, 2024

Abstract With the remarkable progress that technology has made, need for processing data near sensors at edge increased dramatically. The electronic systems used in these applications must process continuously, real-time, and extract relevant information using smallest possible energy budgets. A promising approach implementing always-on of sensory signals supports on-demand, sparse, edge-computing is to take inspiration from biological nervous system. Following this approach, we present a brain-inspired platform prototyping real-time event-based spiking neural networks. system proposed direct emulation dynamic realistic phenomena such as short-term plasticity, NMDA gating, AMPA diffusion, homeostasis, spike frequency adaptation, conductance-based dendritic compartments transmission delays. analog circuits implement primitives are paired with low latency asynchronous digital routing mapping events. This infrastructure enables definition different network architectures, provides interfaces convert encode continuous-signal sensors. Here describe overall architecture, characterize mixed signal analog-digital emulate dynamics, demonstrate their features experimental measurements, low- high-level software ecosystem can be configuring flexibility biologically plausible networks, chip’s ability monitor both population single neuron allow develop validate complex models basic research applications.

Language: Английский

---

Bottom-Up and Top-Down Approaches for the Design of Neuromorphic Processing Systems: Tradeoffs and Synergies Between Natural and Artificial Intelligence

Proceedings of the IEEE, Journal Year: 2023, Volume and Issue: 111(6), P. 623 - 652

Published: June 1, 2023

While Moore's law has driven exponential computing power expectations, its nearing end calls for new avenues improving the overall system performance. One of these is exploration alternative brain-inspired architectures that aim at achieving flexibility and computational efficiency biological neural processing systems. Within this context, neuromorphic engineering represents a paradigm shift in based on implementation spiking network which memory are tightly co-located. In paper, we provide comprehensive overview field, highlighting different levels granularity realized comparing design approaches focus replicating natural intelligence (bottom-up) versus those solving practical artificial applications (top-down). First, present analog, mixed-signal digital circuit styles, identifying boundary between through time multiplexing, in-memory computation, novel devices. Then, highlight key tradeoffs each bottom-up top-down approaches, survey their silicon implementations, carry out detailed comparative analyses to extract guidelines. Finally, identify necessary synergies missing elements required achieve competitive advantage systems over conventional machine-learning accelerators edge applications, outline ingredients framework toward intelligence.

Language: Английский

---

Spike-based local synaptic plasticity: a survey of computational models and neuromorphic circuits

Neuromorphic Computing and Engineering, Journal Year: 2023, Volume and Issue: 3(4), P. 042001 - 042001

Published: Oct. 23, 2023

Abstract Understanding how biological neural networks carry out learning using spike-based local plasticity mechanisms can lead to the development of real-time, energy-efficient, and adaptive neuromorphic processing systems. A large number models have recently been proposed following different approaches. However, it is difficult assess if these be easily implemented in hardware, compare their features ease implementation. To this end, survey, we provide an overview representative brain-inspired synaptic mixed-signal complementary metal–oxide–semiconductor circuits within a unified framework. We review historical, experimental, theoretical approaches modeling plasticity, identify computational primitives that support low-latency low-power hardware implementations rules. common definition locality principle based on pre- postsynaptic signals, which propose as important requirement for physical circuits. Based principle, properties same framework, describe set electronic used implement computing principles, build efficient on-chip online

Language: Английский

---

Memristor-based adaptive neuromorphic perception in unstructured environments

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: May 31, 2024

Abstract Efficient operation of control systems in robotics or autonomous driving targeting real-world navigation scenarios requires perception methods that allow them to understand and adapt unstructured environments with good accuracy, adaptation, generality, similar humans. To address this need, we present a memristor-based differential neuromorphic computing, perceptual signal processing, online adaptation method providing style external sensory stimuli. The ability generality are confirmed two application scenarios: object grasping driving. In the former, robot hand realizes safe stable through fast ( ~ 1 ms) based on tactile features single memristor. latter, decision-making information 10 is extracted an accuracy 94% 40×25 memristor array. By mimicking human low-level mechanisms, electronic circuit-based achieves real-time high-level reactions environments.

Language: Английский

---

Brain-Inspired Computing: A Systematic Survey and Future Trends

Proceedings of the IEEE, Journal Year: 2024, Volume and Issue: 112(6), P. 544 - 584

Published: June 1, 2024

Language: Английский

---

BiœmuS: A new tool for neurological disorders studies through real-time emulation and hybridization using biomimetic Spiking Neural Network

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: June 20, 2024

Abstract Characterization and modeling of biological neural networks has emerged as a field driving significant advancements in our understanding brain function related pathologies. As today, pharmacological treatments for neurological disorders remain limited, pushing the exploration promising alternative approaches such electroceutics. Recent research bioelectronics neuromorphic engineering have fostered development new generation neuroprostheses repair. However, achieving their full potential necessitates deeper biohybrid interaction. In this study, we present novel real-time, biomimetic, cost-effective user-friendly network capable real-time emulation experiments. Our system facilitates investigation replication biophysically detailed dynamics while prioritizing cost-efficiency, flexibility ease use. We showcase feasibility conducting experiments using standard biophysical interfaces variety cells well diverse configurations. envision crucial step towards neuromorphic-based bioelectrical therapeutics, enabling seamless communication with on comparable timescale. Its embedded functionality enhances practicality accessibility, amplifying its real-world applications

Language: Английский

---

Advancing brain-inspired computing with hybrid neural networks

National Science Review, Journal Year: 2024, Volume and Issue: 11(5)

Published: Feb. 26, 2024

ABSTRACT Brain-inspired computing, drawing inspiration from the fundamental structure and information-processing mechanisms of human brain, has gained significant momentum in recent years. It emerged as a research paradigm centered on brain–computer dual-driven multi-network integration. One noteworthy instance this is hybrid neural network (HNN), which integrates computer-science-oriented artificial networks (ANNs) with neuroscience-oriented spiking (SNNs). HNNs exhibit distinct advantages various intelligent tasks, including perception, cognition learning. This paper presents comprehensive review an emphasis their origin, concepts, biological perspective, construction framework supporting systems. Furthermore, insights suggestions for potential directions are provided aiming to propel advancement HNN paradigm.

Language: Английский

---

Darwin3: a large-scale neuromorphic chip with a novel ISA and on-chip learning

National Science Review, Journal Year: 2024, Volume and Issue: 11(5)

Published: March 17, 2024

ABSTRACT Spiking neural networks (SNNs) are gaining increasing attention for their biological plausibility and potential improved computational efficiency. To match the high spatial-temporal dynamics in SNNs, neuromorphic chips highly desired to execute SNNs hardware-based neuron synapse circuits directly. This paper presents a large-scale chip named Darwin3 with novel instruction set architecture, which comprises 10 primary instructions few extended instructions. It supports flexible model programming local learning rule designs. The architecture is designed mesh of computing nodes an innovative routing algorithm. We used compression mechanism represent synaptic connections, significantly reducing memory usage. up 2.35 million neurons, making it largest its kind on scale. experimental results showed that code density was by 28.3× Darwin3, core fan-in fan-out were 4096× 3072× connection compared physical depth. Our also provided saving between 6.8× 200.8× when mapping convolutional spiking onto chip, demonstrating state-of-the-art performance accuracy latency other chips.

Language: Английский

---