XFXincheng Feng

Xincheng Feng

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Xincheng Feng has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Hardware-aware Low-Rank Adaptation for Large Language Models Based on Hybrid Compute-in-Memory Architecture

    by Taiqiang Wu, Chenchen Ding, Wenyong Zhou, Yuxin Cheng, Xincheng Feng, Shuqi Wang, Wendong Xu, Chufan Shi, Zhengwu Liu, Ngai Wong

    Low-rank adaptation (LoRA) is a predominant parameter-efficient finetuning method for adapting large language models (LLMs) to downstream tasks. Meanwhile, Compute-in-Memory (CIM) architectures demonstrate superior energy efficiency due to their array-level parallel in-memory computing designs. In this article, we propose deploying the LoRA-finetuned LLMs on the hybrid CIM architecture (i.e., pretrained weights onto energy-efficient Resistive Random-Access Memory (RRAM) and LoRA branches onto noise-free Static Random-Access Memory (SRAM)), reducing the energy cost to about 3% compared with the Nvidia A100 GPU. However, the inherent noise of RRAM on the saved weights leads to performance degradation, simultaneously. To address this issue, we design a novel Hardware-aware Low-rank Adaptation (HaLoRA) method. The key insight is to train a LoRA branch that is robust toward such noise and then deploy it on noise-free SRAM, while the extra cost is negligible since the parameters of LoRAs are much fewer than pretrained weights (e.g., 0.15% for LLaMA-3.2 1B model). To improve the robustness towards the noise, we theoretically analyze the gap between the optimization trajectories of the LoRA branch under both ideal and noisy conditions and further design an extra loss to minimize the upper bound of this gap. Therefore, we can enjoy both energy efficiency and accuracy during inference. Experiments finetuning the Qwen and LLaMA series demonstrate the effectiveness of HaLoRA across multiple reasoning tasks, achieving up to 22.7 improvement in average score while maintaining robustness at various noise types and noise levels.


  • A Wireless Pulse Diagnostic Device Utilizing MEMS Pressure Sensor Array for Sleep Apnea Detection

    by Tianxiang Liang, Zhiming Long, Haoyu Tan, Xin Cheng, Yubo Huang, Cao Xia, Yuanlin Xia, Zhuqing Wang

    Inspired by traditional Chinese medicine pulse diagnosis, a wireless pulse diagnostic device, based on a Micro Electro Mechanical Systems (MEMS) pressure sensor array, was designed to detect sleep apnea. The arrayed MEMS pressure sensor configuration within a silica gel casing, capturing multidimensional pulse signals at the radial artery positions. Pulse signals were collected from volunteers during normal breathing and breath-holding scenarios to establish a comprehensive sleep apnea recognition database. A feature vector incorporating the pulse contour characteristic K-value, Reflection Index (RI), Power Ratio (PR), and Heart Rate (HR) was created for classification model training using a Support Vector Machine (SVM). Experimental results showed that the accuracy of pulse wave signal feature vector classification is 99.33% and the sensitivity is 98%. Overall, this device offers innovative technical support and significant application potential to advance modernized traditional Chinese diagnostic methods and intelligent medical devices.


  • Open Access

    Stochastic Multivariate Universal-Radix Finite-State Machine: a Theoretically and Practically Elegant Nonlinear Function Approximator

    by Xincheng Feng, Guodong Shen, Jianhao Hu, Meng Li, Ngai Wong

    Nonlinearities are crucial for capturing complex input-output relationships especially in deep neural networks. However, nonlinear functions often incur various hardware and compute overheads. Meanwhile, stochastic computing (SC) has emerged as a promising approach to tackle this challenge by trading output precision for hardware simplicity. To this end, this paper proposes a first-of-its-kind stochastic multivariate universal-radix finite-state machine (SMURF) that harnesses SC for hardware-simplistic multivariate nonlinear function generation at high accuracy. We present the finite-state machine (FSM) architecture for SMURF, as well as analytical derivations of sampling gate coefficients for accurately approximating generic nonlinear functions. Experiments demonstrate the superiority of SMURF, requiring only 16.07% area and 14.45% power consumption of Taylor-series approximation, and merely 2.22% area of look-up table (LUT) schemes.