Article
Optics
Krzysztof Tyszka, Magdalena Furman, Rafal Mirek, Mateusz Krol, Andrzej Opala, Bartlomiej Seredynski, Jan Suffczynski, Wojciech Pacuski, Michal Matuszewski, Jacek Szczytko, Barbara Pietka
Summary: This paper introduces a new approach to neuromorphic photonics using microcavities as building blocks of optical spiking neurons, and demonstrates its feasibility through experimental results. The research shows that exciton-polaritons exhibit characteristics similar to the Leaky Integrate-and-Fire (LIF) mechanism, enabling ultrafast processing on sub-ns timescales.
LASER & PHOTONICS REVIEWS
(2023)
Article
Neurosciences
Pablo Stoliar, Olivier Schneegans, Marcelo J. Rozenberg
Summary: The study demonstrates the use of ultra-compact neurons to build a functional spiking neural network, providing a new platform for neuroscientific research.
FRONTIERS IN NEUROSCIENCE
(2021)
Article
Neurosciences
Sijia Lu, Feng Xu
Summary: This paper establishes a precise mathematical mapping between the biological parameters of the LIF/SNNs and the parameters of ReLU-AN/DNNs, which is validated through simulation and experiments.
FRONTIERS IN NEUROSCIENCE
(2022)
Article
Computer Science, Artificial Intelligence
Zhenzhi Wu, Hehui Zhang, Yihan Lin, Guoqi Li, Meng Wang, Ye Tang
Summary: The LIAF-Net introduces a new neuron model and deep network structure that efficiently integrates analog values for spatiotemporal processing and avoids the performance loss seen in traditional LIF-SNNs.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Jeeson Kim, Vladmir Kornijcuk, Changmin Ye, Doo Seok Jeong
Summary: This study presents a method to emulate a leaky integrate-and-fire (LIF) model in a hardware-efficient manner using a simplified spike-response model (SRM0) and the template-scaling-based exponential function approximation (TS-EFA). The implementation in FPGA of 512 neurons conforming to SRM0 showcases high precision, low latency, and efficient hardware usage.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2021)
Article
Multidisciplinary Sciences
Verena Brehm, Johannes W. Austefjord, Serban Lepadatu, Alireza Qaiumzadeh
Summary: The proposed neuromorphic computing model based on antiferromagnetic domain walls can mimic the behavior of biological neurons and has faster processing speed and more functionalities compared to previous models based on ferromagnetic systems.
SCIENTIFIC REPORTS
(2023)
Article
Computer Science, Theory & Methods
Ioannis E. Venetis, Astero Provata
Summary: This study analyzes the performance issues of implementing the coupled Leaky Integrate-and-Fire model on a GPU, finding that the problem is mainly memory-bound. The results demonstrate that using advanced memory technology on a GPU can achieve better performance.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2022)
Article
Chemistry, Analytical
Malik Summair Asghar, Saad Arslan, Hyungwon Kim
Summary: This study presents an area and power optimized hardware implementation of a large-scale Spiking Neural Network for real time IoT applications. The optimization of neuron and synapse circuits for input image and power consumption reduction is highlighted.
Article
Chemistry, Analytical
Malik Summair Asghar, Saad Arslan, Ali A. A. Al-Hamid, HyungWon Kim
Summary: This paper presents a compact analog SoC implementation of a low-power spiking neural network (SNN) for IoT applications. The SNN is composed of optimized analog neuron and synapse circuits. The proposed synapse circuit based on CMCI provides reduced power consumption and chip area and scalability for higher resolutions. The asynchronous structure of the proposed neuron circuit enables higher energy efficiency. Compared to a digital counterpart, the implemented SNN chip achieves a significantly lower power consumption and occupies a smaller chip area.
Article
Computer Science, Artificial Intelligence
Changqing Xu, Yi Liu, Yintang Yang
Summary: In this work, a compression method is proposed to aggregate individual events into a few time steps of synaptic current, reducing training and inference latency. Multiple-threshold Leaky Integrate-and-Fire (LIF) models with learnable membrane time constants are introduced to enhance information processing capability. Experimental results demonstrate that the proposed method achieves higher accuracy compared to state-of-the-art approaches.
Article
Computer Science, Artificial Intelligence
Parvaneh Rashvand, Mohammad Reza Ahmadzadeh, Farzaneh Shayegh
Summary: Spiking neural networks (SNNs) work based on temporal coding approaches and can achieve a more robust network by optimizing parameters and increasing information transmission. A robust SNN based on the Integrate-and-Fire neuron model and a new training method was proposed and evaluated on multiple datasets in this paper. Additionally, the proposed SNN was applied to the ABIDE1 dataset and achieved a good accuracy rate.
INTERNATIONAL JOURNAL OF NEURAL SYSTEMS
(2021)
Article
Engineering, Electrical & Electronic
Ajay Kumar Singh, Vivek Saraswat, Maryam Shojaei Baghini, Udayan Ganguly
Summary: This study demonstrates the hardware implementation of a recurrent SNN using low-power, low-area, and low-leakage band-to-band-tunneling (BTBT) based neurons. The proposed neurons achieve significant improvements in energy efficiency and standby power compared to existing architectures, enabling brain scale computing with similar area requirements.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2022)
Article
Physics, Applied
Xiangyu Chen, Takeaki Yajima, Isao H. Inoue, Tetsuya Iizuka
Summary: This article proposes a compact leaky integrate-and-fire (LIF) neuron circuit with a long and tunable time constant, providing a solution for the large-scale integration of adaptive SNNs.
JAPANESE JOURNAL OF APPLIED PHYSICS
(2022)
Article
Neurosciences
Youngeun Kim, Yuhang Li, Abhishek Moitra, Ruokai Yin, Priyadarshini Panda
Summary: Spiking Neural Networks (SNNs) are energy efficient neural networks that have gained attention for their binary and asynchronous computation. However, the non-linear activation of LIF neurons in SNNs requires additional memory. This study proposes EfficientLIF-Net, a solution that shares LIF neurons across different layers and channels, achieving comparable accuracy while improving memory efficiency for LIF neurons.
FRONTIERS IN NEUROSCIENCE
(2023)
Article
Engineering, Electrical & Electronic
Avinash Lahgere
Summary: In this article, a trench Bipolar I-MOS is used for the realization of a spiking neural network by using calibrated 2-D TCAD simulations. The proposed trench Bipolar I-MOS LIF neuron can emulate the biological neuronal nature and has a low threshold voltage (-0.16 V) lower than the past reported LBIMOS LIF neuron. The trench Bipolar I-MOS neuron consumes 0.35 pJ energy per spike, which is much lower compared to other types of LIF neurons.
IEEE TRANSACTIONS ON NANOTECHNOLOGY
(2023)
Article
Multidisciplinary Sciences
Panagiotis S. Ioannou, Evripides Kyriakides, Olivier Schneegans, John Giapintzakis
SCIENTIFIC REPORTS
(2020)
Article
Neurosciences
Pablo Stoliar, Olivier Schneegans, Marcelo J. Rozenberg
Summary: The study demonstrates the use of ultra-compact neurons to build a functional spiking neural network, providing a new platform for neuroscientific research.
FRONTIERS IN NEUROSCIENCE
(2021)
Article
Multidisciplinary Sciences
Pavel Salev, Lorenzo Fratino, Dayne Sasaki, Rani Berkoun, Javier del Valle, Yoav Kalcheim, Yayoi Takamura, Marcelo Rozenberg, Ivan K. Schuller
Summary: The authors investigate a unique type of metal-to-insulator resistive switching in (La,Sr)MnO3, where an insulating barrier perpendicular to the current is formed. By electrically inducing a transverse barrier, a novel approach to voltage-controlled magnetism is demonstrated, allowing for local on/off control of ferromagnetism.
NATURE COMMUNICATIONS
(2021)
Article
Physics, Applied
P. Stoliar, O. Schneegans, M. J. Rozenberg
Summary: This study investigates the minimal recurrent spiking neural network of a single neuron with an autaptic synapse, implemented in solid state using an ultracompact neuron model based on the memristive properties of a thyristor. By controlling feedback, they explore the systematic behavior and dynamic memory of the network, replicating experimentally observed behavior of biological autapse. This work lays the foundation for solid-state neuroscience research using the ultracompact neuron as a platform.
PHYSICAL REVIEW APPLIED
(2021)
Article
Instruments & Instrumentation
Huy Mai, Alexandre Jaffre, Khai M. Doan, Thien D. Trinh, Olivier Schneegans
Summary: In this paper, a new inverse analytical method is proposed to extract the optical parameters of a slab without the need for numerical iteration processes. The high accuracy of this method is evaluated by determining the optical parameters of CaF2 and Si substrates in the IR spectral range of 4-8 μm.
APPLIED SPECTROSCOPY
(2022)
Article
Physics, Multidisciplinary
Coline Adda, Min-Han Lee, Yoav Kalcheim, Pavel Salev, Rodolfo Rocco, Nicolas M. Vargas, Nareg Ghazikhanian, Chung-Pang Li, Grant Albright, Marcelo Rozenberg, Ivan K. Schuller
Summary: This study demonstrates the resistive switching and oscillatory spiking regimes in V3O5 devices at room temperature and provides insights into the underlying mechanisms. The findings highlight the potential of V3O5 as a key vanadium oxide for neuromorphic computing.
Article
Physics, Applied
Rodolfo Rocco, Javier del Valle, Henry Navarro, Pavel Salev, Ivan K. Schuller, Marcelo Rozenberg
Summary: Mott materials allow the development of compact and power-efficient neuromorphic devices known as Mott neurons. However, the nature of the insulator-to-metal transition and the determination of the threshold voltage needed for the transition have not been fully understood. In this study, numerical simulations and experiments are used to investigate the filament incubation and formation process. The results show that both electronic and thermal effects contribute to filamentary growth, and the percolation of metallic filaments near the threshold exhibits stochastic behavior.
PHYSICAL REVIEW APPLIED
(2022)
Article
Physics, Multidisciplinary
P. Stoliar, I Akita, O. Schneegans, M. Hioki, M. J. Rozenberg
Summary: This study presents a VLSI implementation of a neuron based on a Silicon-Controlled Rectifier (SCR), which possesses the functionality of the leaky-integrate and fire model (LIF) of spiking neurons. To efficiently migrate the SCR to VLSI, a MOS transistor-based circuit is proposed as an alternative. The results of Spice simulation, as well as VLSI layout and post layout simulations for a 65 nm CMOS process, validate the feasibility of this approach.
JOURNAL OF PHYSICS COMMUNICATIONS
(2022)
Article
Physics, Applied
C. Adda, H. Navarro, J. Kaur, M. -H. Lee, C. Chen, M. Rozenberg, S. P. Ong, Ivan K. Schuller
Summary: This study reports an optically induced resistive switching device based on a CdS/V3O5 heterostructure, which enables nonvolatile resistive switching at room temperature. By transferring photoinduced carriers from CdS to V3O5 under illumination, controllable resistance state switching is achieved. This system has significant potential for various applications.
APPLIED PHYSICS LETTERS
(2022)
Article
Nanoscience & Nanotechnology
Ngoc-Anh Nguyen, Olivier Schneegans, Raphael Salot, Yann Lamy, John Giapintzakis, Van Huy Mai, Sami Oukassi
Summary: This study presents a new solid-state SynT based on LixTiO2 that overcomes the limitations of traditional SynTs, showing excellent endurance, recognition accuracy, and ultralow switching energy. Reversible lithium intercalation enables nonvolatile conductance modulation, while comprehensive electrochemical study provides insight into the specific mechanism of conductance modulation. These results highlight the high potential of LixTiO2-based SynTs for energy-efficient neuromorphic applications.
ADVANCED ELECTRONIC MATERIALS
(2022)
Article
Physics, Applied
Erbin Qiu, Pavel Salev, Lorenzo Fratino, Rodolfo Rocco, Henry Navarro, Coline Adda, Junjie Li, Min-Han Lee, Yoav Kalcheim, Marcelo Rozenberg, Ivan K. Schuller
Summary: In this study, we observed the emergence of an unusual stochastic pattern in coupled spiking Mott nanodevices. We found that increasing the coupling strength leads to stochastic disruptions of the alternating spiking sequence, which is counterintuitive. These disruptions are caused by the small intrinsic stochasticity in electrical triggering of the insulator-metal transition. The stochastic spiking pattern in Mott nanodevices resembles those in biological neurons and has implications for probabilistic computing and biologically plausible electronic devices.
APPLIED PHYSICS LETTERS
(2023)
Article
Materials Science, Multidisciplinary
Qikai Guo, Cesar Magen, Marcelo J. Rozenberg, Beatriz Noheda
Summary: Efforts to understand metallic behavior have led to important concepts, but a unified description of metallic resistivity is still missing. An empirical analysis shows that the parallel resistor formalism used in cuprates can provide a phenomenological description of electrical resistivity in all metals. Two terms of the Taylor expansion of resistivity, detached from their physics origin, are shown to correspond to the T-linear and T-quadratic dependence of electron scattering rates.
Article
Materials Science, Multidisciplinary
L. Fratino, S. Bag, A. Camjayi, M. Civelli, M. Rozenberg
Summary: The resistive collapse of the Mott insulator state in the dimer Hubbard model was studied, revealing the existence of an intermediate bad metallic phase with exotic features such as a pseudogap, orbital selectivity, and a first-order metal-metal transition.