arXiv subjects

Sean Longyu Ma

Publications and source records attributed to Sean Longyu Ma.

6 recordsLinked to original sources

arXiv2025

A Novel Computing Paradigm for MobileNetV3 using Memristor

The increasing computational demands of deep learning models pose significant challenges for edge devices. To address this, we propose a memristor-based circuit design for MobileNetV3, specifically for image classification tasks. Our design leverages the low power consumption and high integration density of memristors, making it suitable for edge computing. The architecture includes optimized memristive convolutional modules, batch normalization modules, activation function modules, global average pooling modules, and fully connected modules. Experimental results on the CIFAR-10 dataset show that our memristor-based MobileNetV3 achieves over 90% accuracy while significantly reducing inference time and energy consumption compared to traditional implementations. This work demonstrates the potential of memristor-based designs for efficient deployment of deep learning models in resource-constrained environments.

cs.AR↗

arXiv2024

Edge AI Empowered Physical Layer Security for 6G NTN: Potential Threats and Future Opportunities

Due to the enormous potential for economic profit offered by artificial intelligence (AI) servers, the field of cybersecurity has the potential to emerge as a prominent arena for competition among corporations and governments on a global scale. One of the prospective applications that stands to gain from the utilization of AI technology is the advancement in the field of cybersecurity. To this end, this paper provides an overview of the possible risks that the physical layer may encounter in the context of 6G Non-Terrestrial Networks (NTN). With the objective of showcasing the effectiveness of cutting-edge AI technologies in bolstering physical layer security, this study reviews the most foreseeable design strategies associated with the integration of edge AI in the realm of 6G NTN. The findings of this paper provide some insights and serve as a foundation for future investigations aimed at enhancing the physical layer security of edge servers/devices in the next generation of trustworthy 6G telecommunication networks.

cs.NI↗

arXiv2024

Empirical Risk-aware Machine Learning on Trojan-Horse Detection for Trusted Quantum Key Distribution Networks

Quantum key distribution (QKD) is a cryptographic technique that leverages principles of quantum mechanics to offer extremely high levels of data security during transmission. It is well acknowledged for its capacity to accomplish provable security. However, the existence of a gap between theoretical concepts and practical implementation has raised concerns about the trustworthiness of QKD networks. In order to mitigate this disparity, we propose the implementation of risk-aware machine learning techniques that present risk analysis for Trojan-horse attacks over the time-variant quantum channel. The trust condition presented in this study aims to evaluate the offline assessment of safety assurance by comparing the risk levels between the recommended safety borderline. This assessment is based on the risk analysis conducted. Furthermore, the proposed trustworthy QKD scenario demonstrates its numerical findings with the assistance of a state-of-the-art point-to-point QKD device, which operates over optical quantum channels spanning distances of 1m, 1km, and 30km. Based on the results from the experimental evaluation of a 30km optical connection, it can be concluded that the QKD device provided prior information to the proposed learner during the non-existence of Eve's attack. According to the optimal classifier, the defensive gate offered by our learner possesses the capability to identify any latent Eve attacks, hence effectively mitigating the risk of potential vulnerabilities. The Eve detection probability is provably bound for our trustworthy QKD scenario.

quant-ph↗

arXiv2024

An Edge AI System Based on FPGA Platform for Railway Fault Detection

As the demands for railway transportation safety increase, traditional methods of rail track inspection no longer meet the needs of modern railway systems. To address the issues of automation and efficiency in rail fault detection, this study introduces a railway inspection system based on Field Programmable Gate Array (FPGA). This edge AI system collects track images via cameras and uses Convolutional Neural Networks (CNN) to perform real-time detection of track defects and automatically reports fault information. The innovation of this system lies in its high level of automation and detection efficiency. The neural network approach employed by this system achieves a detection accuracy of 88.9%, significantly enhancing the reliability and efficiency of detection. Experimental results demonstrate that this FPGA-based system is 1.39* and 4.67* better in energy efficiency than peer implementation on the GPU and CPU platform, respectively.

cs.CV↗

arXiv2024

A RISC-V SOC for Terahertz IoT Devices: Implementation and design challenges

Terahertz (THz) communication is considered a viable approach to augmenting the communication capacity of prospective Internet-of-Things (IoT) resulting in enhanced spectral efficiency. This study first provides an outline of the design challenges encountered in developing THz transceivers. This paper introduces advanced approaches and a unique methodology known as Modified Pulse-width Modulation (MPWM) to address the issues in the THz domain. In this situation involving a transceiver that handles complex modulation schemes, the presence of a mixed signal through a high-resolution digital-to-analog converter (DAC) in the transmitter greatly contributes to the limitation in maintaining linearity at high frequencies. The utilization of Pulse-width Modulation-based Digital-to-Analog Converters (PWM-DACs) has garnered significant attention among scholars due to its efficiency and affordability. However, the converters' performance is restricted by insufficient conversion speed and precision, especially in the context of high-resolution, high-order modulation schemes for THz wireless communications. The MPWM framework offers a multitude of adjustable options, rendering the final MPWM-DAC highly adaptable for a diverse array of application scenarios. Comparative performance assessments indicate that MPWM-DACs have enhanced conversion speed compared to standard PWM-DACs, and they also provide greater accuracy in comparison to Pulse-count Modulation DACs (PCM-DACs). The study presents a comprehensive examination of the core principles, spectrum characteristics, and evaluation metrics, as well as the development and experimental validation of the MPWM method. Furthermore, we present a RISC-V System-on-Chip (SoC) that incorporates an MPWM-DAC, offering a highly favorable resolution for THz IoT communications.

cs.AR↗

arXiv2023

Joint Source-Channel Coding System for 6G Communication: Design, Prototype and Future Directions

The goal of semantic communication is to surpass optimal Shannon's criterion regarding a notable problem for future communication which lies in the integration of collaborative efforts between the intelligence of the transmission source and the joint design of source coding and channel coding. The convergence of scholarly investigation and applicable products in the field of semantic communication is facilitated by the utilization of flexible structural hardware design, which is constrained by the computational capabilities of edge devices. This characteristic represents a significant benefit of joint source-channel coding (JSCC), as it enables the generation of source alphabets with diverse lengths and achieves a code rate of unity. Moreover, JSCC exhibits near-capacity performance while maintaining low complexity. Therefore, we leverage not only quasi-cyclic (QC) characteristics to propose a QC-LDPC code-based JSCC scheme but also Unequal Error Protection (UEP) to ensure the recovery of semantic importance. In this study, the feasibility for using a semantic encoder/decoder that is aware of UEP can be explored based on the existing JSCC system. This approach is aimed at protecting the significance of semantic task-oriented information. Additionally, the deployment of a JSCC system can be facilitated by employing Low-Density Parity-Check (LDPC) codes on a reconfigurable device. This is achieved by reconstructing the LDPC codes as QC-LDPC codes. The QC-LDPC layered decoding technique, which has been specifically optimized for hardware parallelism and tailored for channel decoding applications, can be suitably adapted to accommodate the JSCC system. The performance of the proposed system is evaluated by conducting BER measurements using both floating-point and 6-bit quantization.

cs.IT↗