Nhan Trong Luu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Nhan Trong Luu has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Duong Trung Luu
Can Tho University
7 shared publications - Truong Cong Thang
Department of Computer Science and Engineering, the University of Aizu, Aizuwakamatsu, Japan
6 shared publications - Pham Ngoc Nam
College of Engineering and Computer Science, VinUniversity, Hanoi, Vietnam
5 shared publications Duong Trung Luu
Center for Digital Transformation and Communication, Can Tho University, Can Tho, Vietnam
2 shared publicationsLuu Trung Duong
Center of Digital Transformation and Communication Can Tho University Vietnam
2 shared publications- Jiangkun Wang
The University of Aizu
1 shared publication
Latest Publications
- 4 November 2026
- Published by Springer Nature
- p. 32-47
CEDN: Training Spiking and Hybrid Neural Networks With Constrained Encode-Decode Neurons
- 4 November 2026
- Published by Springer Nature
- p. 32-47
CEDN: Training Spiking and Hybrid Neural Networks With Constrained Encode-Decode Neurons
- 28 October 2026
- Published by Springer Nature in Applied Intelligence
- Vol. 56 (16) , 526
Born-rule physics-informed spiking neural networks for shot-noise-aware quantum state tomography
- 28 October 2026
- Published by Springer Nature in Applied Intelligence
- Vol. 56 (16) , 526
Born-rule physics-informed spiking neural networks for shot-noise-aware quantum state tomography
- 11 October 2026
- Published by Springer Nature
- p. 15-26
HIQI-MLP: Hybrid Interlaced Quantum-Izhikevich Multilayer Perceptron
- 11 October 2026
- Published by Springer Nature
- p. 15-26
HIQI-MLP: Hybrid Interlaced Quantum-Izhikevich Multilayer Perceptron
- 11 October 2026
- Published by Springer Nature
- p. 90-101
FLARE: Forward-only Local Learning via Anticipatory Residual Eligibility for Spiking Neural Networks
- 11 October 2026
- Published by Springer Nature
- p. 90-101
FLARE: Forward-only Local Learning via Anticipatory Residual Eligibility for Spiking Neural Networks
- Open Access
- 25 August 2026
- Published by IOP Publishing in Neuromorphic Computing and Engineering
- Vol. 6 (3) , 034016
Why temporal spike order reversal drops spiking network accuracy and how to partially mitigate it
Spiking neural networks (SNNs) offer an energy-efficient alternative to traditional neural networks by utilizing discrete, temporally precise spike events. However, this study identifies a critical vulnerability in SNNs on recently established bit-based codes: consistent performance degradation when temporal spike encoding orders are reversed, such as using least-significant-bit ordering instead of most-significant-bit. We theoretically formalize this phenomenon as premature state annihilation, wherein early noisy spikes in information-discordant encodings trigger hard resets in leaky integrate-and-fire (LIF) neurons. These resets erase accumulated membrane state and, because the effective temporal influence of an input is largest for early timesteps, leave the backpropagated learning signal concentrated where the information is not. We measure the per-timestep class-mutual-information profile of six encodings directly, without reference to network accuracy, and show that the resulting concordance ordering predicts the observed degradation. While dense codes like weighted phase encoding suffer catastrophic drops (up to 48 % ), sparse codes like time-to-first-spike remain robust, and exchangeable rate codes are provably invariant to reversal. We evaluate several mitigation strategies, finding that parametric LIF (PLIF) neurons and aggressive membrane leakage significantly recover performance by adapting to or suppressing early noise.
- 25 August 2026
- Published by IOP Publishing in Neuromorphic Computing and Engineering
- Vol. 6 (3) , 034016
Open AccessWhy temporal spike order reversal drops spiking network accuracy and how to partially mitigate it
Spiking neural networks (SNNs) offer an energy-efficient alternative to traditional neural networks by utilizing discrete, temporally precise spike events. However, this study identifies a critical vulnerability in SNNs on recently established bit-based codes: consistent performance degradation when temporal spike encoding orders are reversed, such as using least-significant-bit ordering instead of most-significant-bit. We theoretically formalize this phenomenon as premature state annihilation, wherein early noisy spikes in information-discordant encodings trigger hard resets in leaky integrate-and-fire (LIF) neurons. These resets erase accumulated membrane state and, because the effective temporal influence of an input is largest for early timesteps, leave the backpropagated learning signal concentrated where the information is not. We measure the per-timestep class-mutual-information profile of six encodings directly, without reference to network accuracy, and show that the resulting concordance ordering predicts the observed degradation. While dense codes like weighted phase encoding suffer catastrophic drops (up to 48 % ), sparse codes like time-to-first-spike remain robust, and exchangeable rate codes are provably invariant to reversal. We evaluate several mitigation strategies, finding that parametric LIF (PLIF) neurons and aggressive membrane leakage significantly recover performance by adapting to or suppressing early noise.
- 11 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 2335-2339
Direct-to-Event Spiking Neural Network Transfer
Spiking Neural Networks (SNNs) have gained increasing attention due to their potential for low-power computation on neuromorphic hardware. A widely adopted training strategy for SNNs is direct coding, which enable backpropagation on neuron implementations using continuous-valued surrogate activations. However, recent studies have shown that direct-coded SNNs remain substantially less energy-efficient than their event based counterparts, limiting their practical deployment in energy sensitive scenarios. Still, to promote the reusability of pretrained SNN database on direct code, this motivates an important yet underexplored question: How can a SNN pretrained with direct code be effectively converted into an event-based representation? In this research, we present the first systematic investigation into this transfer problem, analyze the key challenges that arise when transitioning from direct-coded to event-based computation and propose a set of methods to enable energy-efficient transfer while preserving model performance.
- 11 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 2335-2339
Direct-to-Event Spiking Neural Network Transfer
Spiking Neural Networks (SNNs) have gained increasing attention due to their potential for low-power computation on neuromorphic hardware. A widely adopted training strategy for SNNs is direct coding, which enable backpropagation on neuron implementations using continuous-valued surrogate activations. However, recent studies have shown that direct-coded SNNs remain substantially less energy-efficient than their event based counterparts, limiting their practical deployment in energy sensitive scenarios. Still, to promote the reusability of pretrained SNN database on direct code, this motivates an important yet underexplored question: How can a SNN pretrained with direct code be effectively converted into an event-based representation? In this research, we present the first systematic investigation into this transfer problem, analyze the key challenges that arise when transitioning from direct-coded to event-based computation and propose a set of methods to enable energy-efficient transfer while preserving model performance.
- Open Access
- 21 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 78497-78557
A Survey on Spiking Neural Network Foundation and Recent Progress
Over the past decade, deep neural networks (DNNs) have achieved remarkable success across numerous fields. Despite this progress, DNNs remain highly demanding in terms of computational power, energy consumption, and data requirements. As the demand for autonomous systems such as self-driving cars, drones, and collaborative robots, etc. continues to rise, the deployment of DNNs in these real-world scenarios has gained significant attention. However, these applications require energy-efficient and computationally lightweight models to ensure real-time performance under limited power resources. A promising alternative has emerged in the form of biologically inspired spiking neural networks (SNNs), which offer a more energy-efficient and temporally dynamic computing paradigm. By mimicking the behavior of biological neurons, SNNs provide a bridge between neuroscience and machine learning, leveraging neural sparsity and temporal coding mechanisms inherent to biological systems. In this work, we make the following contributions: provide a comprehensive overview of biological neuron theories; summarize existing spike-based neuron models from neuroscience; describe synaptic modeling approaches; review conventional artificial neural networks; outline several methodologies for preprocessing, training and conversion approaches for spike-based models; explore SNN applications in various fields; discuss major frameworks supporting SNN implementation.
- 21 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 78497-78557
Open AccessA Survey on Spiking Neural Network Foundation and Recent Progress
Over the past decade, deep neural networks (DNNs) have achieved remarkable success across numerous fields. Despite this progress, DNNs remain highly demanding in terms of computational power, energy consumption, and data requirements. As the demand for autonomous systems such as self-driving cars, drones, and collaborative robots, etc. continues to rise, the deployment of DNNs in these real-world scenarios has gained significant attention. However, these applications require energy-efficient and computationally lightweight models to ensure real-time performance under limited power resources. A promising alternative has emerged in the form of biologically inspired spiking neural networks (SNNs), which offer a more energy-efficient and temporally dynamic computing paradigm. By mimicking the behavior of biological neurons, SNNs provide a bridge between neuroscience and machine learning, leveraging neural sparsity and temporal coding mechanisms inherent to biological systems. In this work, we make the following contributions: provide a comprehensive overview of biological neuron theories; summarize existing spike-based neuron models from neuroscience; describe synaptic modeling approaches; review conventional artificial neural networks; outline several methodologies for preprocessing, training and conversion approaches for spike-based models; explore SNN applications in various fields; discuss major frameworks supporting SNN implementation.
- Open Access
- 2 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 51918-51942
Robust Reinforcement Learning Quadcopter Control System for Offshore Cement Wind Turbine Construction
As the global energy sector pivots toward "gigantic" offshore wind structures, the logistical and safety challenges of marine construction have reached a critical bottleneck. This paper pioneers a transformative reinforcement learning (RL) framework designed to automate offshore cement wind turbine construction through a specialized cluster of high-performance quadcopter agents. To eliminate the optimization interference and "negative transfer" typical of multi-task systems, we introduce a modular architecture that decomposes the construction workflow into distinct, high-precision phases: autonomous navigation, cement dispensing, and energy-aware refueling. At the core of our system is the novel application of Joint Spectral Normalization (SN), which stabilizes training by bounding the Lipschitz constants of deep networks, ensuring monotonic policy improvement even in volatile environments. Benchmarking against five state-of-the-art algorithms (SAC, DDPG, TD3, TRPO, and PPO) reveals that our PPO-based SN framework is the only solution capable of consistent convergence. Crucially, while vanilla RL controllers collapse under stochastic disturbances, our Full SN variant maintains mission-critical stability and high convergence rates across extreme Gaussian and uniform observation noise. This study marks the first successful integration of RL-driven aerial robotics into complex industrial construction, offering a robust path forward for civil engineering automation in the world’s most unpredictable marine settings.
- 2 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 51918-51942
Open AccessRobust Reinforcement Learning Quadcopter Control System for Offshore Cement Wind Turbine Construction
As the global energy sector pivots toward "gigantic" offshore wind structures, the logistical and safety challenges of marine construction have reached a critical bottleneck. This paper pioneers a transformative reinforcement learning (RL) framework designed to automate offshore cement wind turbine construction through a specialized cluster of high-performance quadcopter agents. To eliminate the optimization interference and "negative transfer" typical of multi-task systems, we introduce a modular architecture that decomposes the construction workflow into distinct, high-precision phases: autonomous navigation, cement dispensing, and energy-aware refueling. At the core of our system is the novel application of Joint Spectral Normalization (SN), which stabilizes training by bounding the Lipschitz constants of deep networks, ensuring monotonic policy improvement even in volatile environments. Benchmarking against five state-of-the-art algorithms (SAC, DDPG, TD3, TRPO, and PPO) reveals that our PPO-based SN framework is the only solution capable of consistent convergence. Crucially, while vanilla RL controllers collapse under stochastic disturbances, our Full SN variant maintains mission-critical stability and high convergence rates across extreme Gaussian and uniform observation noise. This study marks the first successful integration of RL-driven aerial robotics into complex industrial construction, offering a robust path forward for civil engineering automation in the world’s most unpredictable marine settings.
- Top Cited PapersOpen Access
- 24 February 2026
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e3554
Parameter efficient hybrid spiking-quantum convolutional neural network with surrogate gradient and quantum data-reupload
The rapid advancement of artificial intelligence (AI) and deep learning (DL) has catalyzed the emergence of several optimization-driven subfields, notably neuromorphic computing and quantum machine learning. Leveraging the differentiable nature of hybrid models, researchers have explored their potential to address complex problems through unified optimization strategies. One such development is the Spiking Quantum Neural Network (SQNN), which combines principles from spiking neural networks (SNNs) and quantum computing. However, existing SQNN implementations often depend on pretrained SNNs due to the non-differentiable nature of spiking activity and the limited scalability of current SNN encoders. In this work, we propose a novel architecture—Spiking-Quantum Data Re-upload Convolutional Neural Network (SQDR-CNN)—that enables joint training of convolutional SNNs and quantum circuits within a single backpropagation framework. Unlike its predecessor, SQDR-CNN allow convergence to reasonable performance without the reliance of pretrained spiking encoder and subsetting datasets. We also clarified some theoretical foundations, testing new design using quantum data-reupload with different training algorithm-initialization and evaluate the performance of the proposed model under noisy simulated quantum environments. As a result, we were able to achieve 86% of the mean top-performing accuracy of the state of the art (SOTA) SNN baselines, yet uses only 0.5% of the smallest spiking model’s parameters. Through this integration of neuromorphic and quantum paradigms, we aim to open new research directions and foster technological progress in multi-modal, learnable systems.
- 24 February 2026
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e3554
Top Cited PapersOpen AccessParameter efficient hybrid spiking-quantum convolutional neural network with surrogate gradient and quantum data-reupload
The rapid advancement of artificial intelligence (AI) and deep learning (DL) has catalyzed the emergence of several optimization-driven subfields, notably neuromorphic computing and quantum machine learning. Leveraging the differentiable nature of hybrid models, researchers have explored their potential to address complex problems through unified optimization strategies. One such development is the Spiking Quantum Neural Network (SQNN), which combines principles from spiking neural networks (SNNs) and quantum computing. However, existing SQNN implementations often depend on pretrained SNNs due to the non-differentiable nature of spiking activity and the limited scalability of current SNN encoders. In this work, we propose a novel architecture—Spiking-Quantum Data Re-upload Convolutional Neural Network (SQDR-CNN)—that enables joint training of convolutional SNNs and quantum circuits within a single backpropagation framework. Unlike its predecessor, SQDR-CNN allow convergence to reasonable performance without the reliance of pretrained spiking encoder and subsetting datasets. We also clarified some theoretical foundations, testing new design using quantum data-reupload with different training algorithm-initialization and evaluate the performance of the proposed model under noisy simulated quantum environments. As a result, we were able to achieve 86% of the mean top-performing accuracy of the state of the art (SOTA) SNN baselines, yet uses only 0.5% of the smallest spiking model’s parameters. Through this integration of neuromorphic and quantum paradigms, we aim to open new research directions and foster technological progress in multi-modal, learnable systems.
- Open Access
Parameter efficient hybrid spiking-quantum convolutional neural network with surrogate gradient and quantum data-reupload
The rapid advancement of artificial intelligence (AI) and deep learning (DL) has catalyzed the emergence of several optimization-driven subfields, notably neuromorphic computing and quantum machine learning. Leveraging the differentiable nature of hybrid models, researchers have explored their potential to address complex problems through unified optimization strategies. One such development is the Spiking Quantum Neural Network (SQNN), which combines principles from spiking neural networks (SNNs) and quantum computing. However, existing SQNN implementations often depend on pretrained SNNs due to the non-differentiable nature of spiking activity and the limited scalability of current SNN encoders. In this work, we propose a novel architecture, Spiking-Quantum Data Re-upload Convolutional Neural Network (SQDR-CNN), that enables joint training of convolutional SNNs and quantum circuits within a single backpropagation framework. Unlike its predecessor, SQDR-CNN allow convergence to reasonable performance without the reliance of pretrained spiking encoder and subsetting datasets. We also clarified some theoretical foundations, testing new design using quantum data-reupload with different training algorithm-initialization and evaluate the performance of the proposed model under noisy simulated quantum environments. As a result, we were able to achieve 86% of the mean top-performing accuracy of the SOTA SNN baselines, yet uses only 0.5% of the smallest spiking model's parameters. Through this integration of neuromorphic and quantum paradigms, we aim to open new research directions and foster technological progress in multi-modal, learnable systems.
Open AccessParameter efficient hybrid spiking-quantum convolutional neural network with surrogate gradient and quantum data-reupload
The rapid advancement of artificial intelligence (AI) and deep learning (DL) has catalyzed the emergence of several optimization-driven subfields, notably neuromorphic computing and quantum machine learning. Leveraging the differentiable nature of hybrid models, researchers have explored their potential to address complex problems through unified optimization strategies. One such development is the Spiking Quantum Neural Network (SQNN), which combines principles from spiking neural networks (SNNs) and quantum computing. However, existing SQNN implementations often depend on pretrained SNNs due to the non-differentiable nature of spiking activity and the limited scalability of current SNN encoders. In this work, we propose a novel architecture, Spiking-Quantum Data Re-upload Convolutional Neural Network (SQDR-CNN), that enables joint training of convolutional SNNs and quantum circuits within a single backpropagation framework. Unlike its predecessor, SQDR-CNN allow convergence to reasonable performance without the reliance of pretrained spiking encoder and subsetting datasets. We also clarified some theoretical foundations, testing new design using quantum data-reupload with different training algorithm-initialization and evaluate the performance of the proposed model under noisy simulated quantum environments. As a result, we were able to achieve 86% of the mean top-performing accuracy of the SOTA SNN baselines, yet uses only 0.5% of the smallest spiking model's parameters. Through this integration of neuromorphic and quantum paradigms, we aim to open new research directions and foster technological progress in multi-modal, learnable systems.