Neurobiological systems continually interact with the surrounding environment to refine their behaviour toward the best possible reward. Achieving such learning by experience is one of the main challenges of artificial intelligence, but currently it is hindered by the lack of hardware capable of plastic adaptation. Here, we propose a bio-inspired recurrent neural network, mastered by a digital system on chip with resistive-switching synaptic arrays of memory devices, which exploits homeostatic Hebbian learning for improved efficiency. All the results are discussed experimentally and theoretically, proposing a conceptual framework for benchmarking the main outcomes in terms of accuracy and resilience. To test the proposed architecture for reinforcement learning tasks, we study the autonomous exploration of continually evolving environments and verify the results for the Mars rover navigation. We also show that, compared to conventional deep learning techniques, our in-memory hardware has the potential to achieve a significant boost in speed and power-saving.
Download full-text PDF |
Source |
|---|---|
| http://www.ncbi.nlm.nih.gov/pmc/articles/PMC10030830 | PMC |
| http://dx.doi.org/10.1038/s41467-023-37097-5 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Hierarchically Engineered Self-Adaptive Nanoplatform Guided Intuitive and Precision Interventions for Deep-Seated Glioblastoma.
ACS Nano
January 2025
Shanghai Frontiers Science Center of Drug Target Identification and Delivery, School of Pharmaceutical Sciences, Shanghai Jiao Tong University, Shanghai 200240, China.
Glioblastoma multiforme (GBM), particularly the deep-seated tumor where surgical removal is not feasible, poses great challenges for clinical treatments due to complicated biological barriers and the risk of damaging healthy brain tissue. Here, we hierarchically engineer a self-adaptive nanoplatform (SAN) that overcomes delivery barriers by dynamically adjusting its structure, surface charge, particle size, and targeting moieties to precisely distinguish between tumor and parenchyma cells. We further devise a AN-uided ntuitive and recision ntervention (SGIPi) strategy which obviates the need for sophisticated facilities, skilled operations, and real-time magnetic resonance imaging (MRI) guidance required by current MRI-guided laser or ultrasound interventions.
View Article and Find Full Text PDFStretchable hierarchical metal wire networks for neuromorphic emulation of nociception and anti-nociception.
Mater Horiz
January 2025
Chemistry and Physics of Materials Unit, Jawaharlal Nehru Centre for Advanced Scientific Research, Bangalore 560064, India.
Among biomimetic technologies, the incorporation of sensory hardware holds exceptional utility in human-machine interfacing. In this context, devices receptive to nociception and emulating antinociception gain significance as part of pain management. Here we report, a stretchable two-terminal resistive neuromorphic device consisting of a hierarchical Ag microwire network formed using a crack templating protocol.
View Article and Find Full Text PDFSpQuant-SNN: ultra-low precision membrane potential with sparse activations unlock the potential of on-device spiking neural networks applications.
Front Neurosci
September 2024
School of Electrical and Computer Engineering, Cornell Tech, New York, NY, United States.
Article Synopsis
- Spiking neural networks (SNNs) are gaining popularity for their energy efficiency and biological realism, but they still rely on high-precision values for key components like membrane potential, leading to inefficiencies in resource-limited settings.* -
- Existing approaches to reduce these issues have resulted in significant accuracy drops, particularly when implementing low-precision representation or time step reduction.* -
- The proposed SpQuant-SNN addresses these challenges by integrating an integer-only quantization scheme, spatial-channel pruning, and a self-adaptive learnable threshold to improve performance while maintaining low computational demand and memory usage.*
Spontaneous Threshold Lowering Neuron using Second-Order Diffusive Memristor for Self-Adaptive Spatial Attention.
Adv Sci (Weinh)
August 2023
Department of Electrical and Electronic Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, China.
Intrinsic plasticity of neurons, such as spontaneous threshold lowering (STL) to modulate neuronal excitability, is key to spatial attention of biological neural systems. In-memory computing with emerging memristors is expected to solve the memory bottleneck of the von Neumann architecture commonly used in conventional digital computers and is deemed a promising solution to this bioinspired computing paradigm. Nonetheless, conventional memristors are incapable of implementing the STL plasticity of neurons due to their first-order dynamics.
View Article and Find Full Text PDFA self-adaptive hardware with resistive switching synapses for experience-based neurocomputing.
Nat Commun
March 2023
Dipartimento di Elettronica, Informazione e Bioingegneria, Politecnico di Milano and IUNET, Milano, 20133, Italy.
Neurobiological systems continually interact with the surrounding environment to refine their behaviour toward the best possible reward. Achieving such learning by experience is one of the main challenges of artificial intelligence, but currently it is hindered by the lack of hardware capable of plastic adaptation. Here, we propose a bio-inspired recurrent neural network, mastered by a digital system on chip with resistive-switching synaptic arrays of memory devices, which exploits homeostatic Hebbian learning for improved efficiency.
View Article and Find Full Text PDFWant AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!