Article
Engineering, Multidisciplinary
Saurabh Balkrishna Tandale, Marcus Stoffel
Summary: The present study aims to introduce an AI algorithm suitable for neuromorphic computing to solve Boundary Value Problems in Engineering Mechanics. By using Spiking Neural Networks (SNNs), the study proposes a surrogate model for mechanical tasks that is more energy-efficient than traditional neural networks. The researchers also propose a hybrid model that combines spiking recurrent cells, the spiking variant of the Legendre Memory Unit (LMU), and classical dense transformations to compute the nonlinear response of shock wave-loaded plate elements.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Man Yao, Guangshe Zhao, Hengyu Zhang, Yifan Hu, Lei Deng, Yonghong Tian, Bo Xu, Guoqi Li
Summary: This paper studies the application of attention mechanisms in brain-inspired spiking neural networks (SNNs). By optimizing the membrane potentials using a multi-dimensional attention module, the performance and energy efficiency of SNNs are improved. Experimental results demonstrate that SNNs with attention achieve better performance and sparser spiking firing in event-based action recognition and image classification tasks. The effectiveness of attention SNNs is theoretically proven and further analyzed using a proposed spiking response visualization method. This work highlights the potential of SNNs as a general backbone for various applications in the field of SNN research.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Review
Computer Science, Information Systems
M. Lakshmi Varshika, Federico Corradi, Anup Das
Summary: Sustainable computing requires energy-efficient processors, and neuromorphic systems mimic biological functions for brain-like efficiency, speed, adaptability, and intelligence. Current neuromorphic technology trends focus on novel materials to enable high-integration and extreme low-power brain-inspired computing. Nonvolatile memory technologies are being used for efficient in-memory and in-device computing with spike-based neuromorphic architectures.
Article
Neurosciences
Yijian Pei, Changqing Xu, Zili Wu, Yi Liu, Yintang Yang
Summary: This study proposes an ultra-low latency adaptive local binary spiking neural network (ALBSNN) with accuracy loss estimators, which dynamically selects network layers for binarization to balance quantization degree and classification accuracy. It introduces the global average pooling (GAP) layer and binary weight optimization (BWO) to accelerate network training speed and reduce error caused by binary weights. The combination of these methods enables the network to maintain recognition capability and reduce storage space. The proposed network achieves advanced accuracy in SNN with binary weights, while also having advantages in terms of storage resources and training time.
FRONTIERS IN NEUROSCIENCE
(2023)
Article
Telecommunications
Hyeryung Jang, Nicolas Skatchkovsky, Osvaldo Simeone
Summary: Spiking Neural Networks (SNNs) are biologically inspired machine learning models that process binary and sparse spiking signals. They can be implemented on energy-efficient neuromorphic computing platforms and have been validated for their capabilities in detecting and generating spatial patterns through experiments.
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Computer Science, Artificial Intelligence
Yanhu Chen, Cen Wang, Hongxiang Guo, Xiong Gao, Jian Wu
Summary: This study proposes a quantum algorithm to reduce the complexity of key steps in spiking neural networks (SNNs) and builds a quantum spiking neuron network (QSNN). Experimental results demonstrate the feasibility and robustness of QSNN.
Article
Energy & Fuels
Juan Manuel Gonzalez Sopena, Vikram Pakrashi, Bidisha Ghosh
Summary: This paper proposes a short-term wind power forecasting model based on spiking neural networks adapted to the computational abilities of neuromorphic devices. A case study is conducted using real wind power generation data from Ireland, and the results demonstrate the effectiveness and feasibility of the proposed approach.
Article
Computer Science, Theory & Methods
Zhijie Yang, Lei Wang, Wei Shi, Yao Wang, Junbo Tie, Feng Wang, Xiang Yu, Linghui Peng, Chao Xiao, Xun Xiao, Yao Yao, Gan Zhou, Xuhu Yu, Rui Gong, Xia Zhao, Yuhua Tang, Weixia Xu
Summary: This paper proposes RVNE to enable fine-grained and flexible homogeneous programming for neuromorphic algorithms, and implements a neuromorphic micro-architecture tightly coupled to the CPU pipeline. Evaluation results show significant improvements in code density and energy consumption with RVNE.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Neurosciences
Yang Li, Dongcheng Zhao, Yi Zeng
Summary: This paper introduces the characteristics and problems of spiking neural networks (SNN) in information processing, and proposes a novel bistable spiking neural network (BSNN) to improve the performance of converted SNNs. By designing synchronous neurons (SN), the performance of ANN conversion based on the ResNet structure can be effectively improved. Experimental results show that this method achieves good conversion results on different datasets.
FRONTIERS IN NEUROSCIENCE
(2022)
Article
Nanoscience & Nanotechnology
Hao-Kai Peng, Yu-Kai Huang, Chuan-Pu Chou, Yung-Hsien Wu
Summary: Epitaxial GeSn shows higher polarization capabilities than epi-Ge, and when combined with high-k AION for interface engineering, it enhances the reliability of FeCAPs. The simplified neuromorphic network, functioning as a differentiator circuit, not only performs LIF functions but also recognizes spatiotemporal features efficiently. This energy-efficient network, utilizing FeCAPs, is suitable for emerging spiking neural network hardware accelerators.
ACS APPLIED MATERIALS & INTERFACES
(2021)
Article
Computer Science, Artificial Intelligence
Susanna Gordleeva, Yuliya A. Tsybina, Mikhail I. Krivonosov, Ivan Y. Tyukin, Victor B. Kazantsev, Alexey Zaikin, Alexander N. Gorban
Summary: Mammalian brains can react quickly and effectively to danger due to their ability to recognize patterns. This study proposes a neuron-astrocyte network topology that is adapted to accommodate situation-based memory. Numerical simulations show that astrocytes are necessary for effective function in such a learning and testing set-up. The results also demonstrate that a neuromorphic computational model with astrocyte-induced modulation enhances retrieval quality compared to standard neural networks.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Telecommunications
Nicolas Skatchkovsky, Hyeryung Jang, Osvaldo Simeone
Summary: The synergies between wireless communications and artificial intelligence are increasingly driving research at the intersection of the two fields. Machine learning can address algorithm and model deficits in the optimization of communication protocols, but implementing ML models on devices connected via bandwidth-constrained channels remains challenging.
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Telecommunications
Nicolas Skatchkovsky, Hyeryung Jang, Osvaldo Simeone
Summary: This paper introduces the operation principle, training algorithms, and models of Spiking Neural Networks (SNNs) and compares two main methods. To address the non-differentiability of the spiking mechanism, a differentiable function approximating the threshold activation function is proposed, and an alternative method based on probability models is discussed. Finally, experimental results on accuracy and convergence under different SNN models are provided.
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Computer Science, Hardware & Architecture
Bleema Rosenfeld, Osvaldo Simeone, Bipin Rajendran
Summary: This paper proposes a novel approach for training spiking neural networks (SNNs) by matching distributions of spiking signals instead of individual spiking signals. The approach combines a conditional generator (SNN) and a discriminator (ANN), and utilizes an adversarial iterative learning strategy. The proposed SpikeGAN method is extended with Bayesian learning to capture multi-modal spatio-temporal distribution, and an online meta-learning variant is introduced to handle time-varying statistics. Experiments show the advantages of this approach compared to existing solutions, and demonstrate the use of SpikeGAN to generate neuromorphic datasets for SNN classification.
IEEE TRANSACTIONS ON COMPUTERS
(2022)
Article
Engineering, Biomedical
Yifei Feng, Shijia Geng, Jianjun Chu, Zhaoji Fu, Shenda Hong
Summary: This paper introduces a deep spiking neural network (SNN) for ECG classification, and compares the effects of different ANN activation functions on SNN performance.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2022)
Article
Multidisciplinary Sciences
Mohammadali Sharifshazileh, Karla Burelo, Johannes Sarnthein, Giacomo Indiveri
Summary: The study introduces an embedded device that can process intracranial EEG from epilepsy patients to detect High Frequency Oscillations in real-time. By utilizing mixed-signal neuromorphic circuits, a neural network processing system achieves state-of-the-art accuracy, sensitivity, and specificity in HFO detection.
NATURE COMMUNICATIONS
(2021)
Article
Multidisciplinary Sciences
Chiara Bartolozzi, Giacomo Indiveri, Elisa Donati
Summary: This Perspective discusses the potential, challenges and future direction of research aimed at demonstrating embodied intelligent robotics via neuromorphic technology. Neuromorphic engineering studies neural computational principles to develop compact and low-power processing systems. Endowing robots with neuromorphic technologies represents a promising approach for creating robots that can seamlessly integrate in society.
NATURE COMMUNICATIONS
(2022)
Article
Multidisciplinary Sciences
Karla Burelo, Georgia Ramantani, Giacomo Indiveri, Johannes Sarnthein
Summary: This study analyzed scalp EEG recordings from pediatric focal lesional epilepsy patients and developed a custom SNN to detect events of interest and reject artifacts. The occurrence of HFO detected was associated with active epilepsy and correlated with a decrease in seizure frequency.
SCIENTIFIC REPORTS
(2022)
Article
Nanoscience & Nanotechnology
Mohammad Javad Mirshojaeian Hosseini, Elisa Donati, Giacomo Indiveri, Robert A. Nawrocki
Summary: The study demonstrates a physically flexible organic synaptic circuit fabricated using organic materials that offer advantages in terms of time constants, flexibility, and biocompatibility, capable of emulating the behavior of biological synapses. It shows promising results in terms of time constants before and during bending, indicating potential for applications in the field of neural coding and spatiotemporal pattern encoding.
ADVANCED ELECTRONIC MATERIALS
(2022)
Article
Multidisciplinary Sciences
Giorgia Dellaferrera, Stanislaw Wozniak, Giacomo Indiveri, Angeliki Pantazi, Evangelos Eleftheriou
Summary: The article proposes a brain-inspired optimizer based on mechanisms of synaptic integration and strength regulation for improved performance of both artificial and spiking neural networks.
NATURE COMMUNICATIONS
(2022)
Article
Multidisciplinary Sciences
Rohit Abraham John, Yigit Demirag, Yevhen Shynkarenko, Yuliia Berezovska, Natacha Ohannessian, Melika Payvand, Peng Zeng, Maryna Bodnarchuk, Frank Krumeich, Goekhan Kara, Ivan Shorubalko, Manu Nair, Graham A. Cooke, Thomas Lippert, Giacomo Indiveri, Maksym Kovalenko
Summary: This study presents a reconfigurable halide perovskite nanocrystal memristor that can switch between different modes and has excellent endurance. It addresses the diverse switching requirements of various computing frameworks.
NATURE COMMUNICATIONS
(2022)
Review
Neurosciences
Karla Burelo, Mohammadali Sharifshazileh, Giacomo Indiveri, Johannes Sarnthein
Summary: This article introduces a novel method for automatic HFO detection using spiking neural networks and neuromorphic technology, and validates its high accuracy and specificity in different recording modalities. The research contributes to the real-time detection of HFO using compact and low-power neuromorphic devices, and improves the outcome of epilepsy surgery and treatment.
FRONTIERS IN NEUROSCIENCE
(2022)
Article
Multidisciplinary Sciences
Filippo Moro, Emmanuel Hardy, Bruno Fain, Thomas Dalgaty, Paul Clemencon, Alessio De Pra, Eduardo Esmanhotto, Niccolo Castellani, Francois Blard, Francois Gardien, Thomas Mesquida, Francois Rummens, David Esseni, Jerome Casas, Giacomo Indiveri, Melika Payvand, Elisa Vianello
Summary: This study presents a neuromorphic in-memory event-driven system for real-world object localization applications, which is orders of magnitude more energy efficient than microcontrollers.
NATURE COMMUNICATIONS
(2022)
Article
Multidisciplinary Sciences
Melika Payvand, Filippo Moro, Kumiko Nomura, Thomas Dalgaty, Elisa Vianello, Yoshifumi Nishi, Giacomo Indiveri
Summary: Learning plays a crucial role in creating intelligent machines. This study introduces MEMSORN, an adaptive hardware architecture that incorporates resistive memory (RRAM) to achieve self-organizing spiking recurrent neural network. The utilization of technologically plausible learning rules based on Hebbian and Homeostatic plasticity, derived from statistical measurements of fabricated RRAM-based neurons and synapses, improves the network accuracy by 30% in sequence learning tasks. Furthermore, the comparison with a fully-randomly-set-up spiking recurrent network demonstrates that self-organization can enhance the accuracy by over 15%.
NATURE COMMUNICATIONS
(2022)
Proceedings Paper
Engineering, Electrical & Electronic
Junren Chen, Chenxi Wu, Giacomo Indiveri, Melika Payvand
Summary: This paper analyzes the reliability issues of using memristors in routing crossbar arrays, including resource sharing collisions and undesired pulses from leakage paths. It shows that there is a trade-off between receiver connectivity and routing collision probability, and provides specifications and guidelines for engineering memristor devices and designing routing systems.
2022 29TH IEEE INTERNATIONAL CONFERENCE ON ELECTRONICS, CIRCUITS AND SYSTEMS (IEEE ICECS 2022)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Nik Dennler, Germain Haessig, Matteo Cartiglia, Giacomo Indiveri
Summary: This paper proposes a neuromorphic approach using spiking neural networks for vibration analysis, optimized for various scenarios and smaller-scale applications. The method operates online in an unsupervised manner, detecting system anomalies using a cochlea model, feedback adaptation, and a balanced spiking neural network. The proposed method demonstrates state-of-the-art performance on two publicly available datasets and is implemented on an asynchronous neuromorphic processor device, marking progress towards autonomous low-power edge-computing devices for online vibration monitoring.
2021 IEEE 3RD INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE CIRCUITS AND SYSTEMS (AICAS)
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Julian Buechel, Jonathan Kakon, Michel Perez, Giacomo Indiveri
Summary: The article introduces a novel local learning rule suitable for on-chip implementation in low-power neuromorphic processors to drive a randomly connected network of spiking neurons into a tightly balanced regime. The proposed method paves the way towards a system-level implementation of tightly balanced networks on analog mixed-signal neuromorphic hardware.
2021 IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS (ISCAS)
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Nicoletta Risi, Enrico Calabrese, Giacomo Indiveri
Summary: The stereo-matching problem efficiently solved in biology remains a computational bottleneck for classical machine vision approaches. Recently proposed Spiking Neural Network architectures have the potential of simplifying this problem by exploiting the properties of event cameras.
2021 IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS (ISCAS)
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Yigit Demirag, Filippo Moro, Thomas Dalgaty, Gabriele Navarro, Charlotte Frenkel, Giacomo Indiveri, Elisa Vianello, Melika Payvand
Summary: Research on dedicated hardware implementations of spiking neural networks combining advantages of neuromorphic circuits with emerging memory technologies shows potential for ultra-low power sensory processing. A new spike-based learning rule has been proposed to solve the temporal credit assignment problem and approximate the error back-propagation algorithm on complex tasks. The PCM-trace, a new neuromorphic building block leveraging phasechange materials, improves area efficiency and supports a technologically feasible learning algorithm backed by experimental data.
2021 IEEE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS (ISCAS)
(2021)
Proceedings Paper
Computer Science, Theory & Methods
Elisa Donati, Renate Krause, Giacomo Indiveri
Summary: Neuromorphic electronic circuits are ideal for developing brain-inspired low-power implantable processing systems that can interface with biological systems in real-time. These circuits are suitable for continuously monitoring physiological parameters of the body and generating patterns in real-time.
2021 10TH INTERNATIONAL IEEE/EMBS CONFERENCE ON NEURAL ENGINEERING (NER)
(2021)