Arrow Research search

Author name cluster

Liang Zhou

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

8 papers
2 author rows

Possible papers

8

AAAI Conference 2026 Conference Paper

Certified L2-Norm Robustness of 3D Point Cloud Recognition in the Frequency Domain

  • Liang Zhou
  • Qiming Wang
  • Tianze Chen

3D point cloud classification is a fundamental task in safety-critical applications such as autonomous driving, robotics, and augmented reality. However, recent studies reveal that point cloud classifiers are vulnerable to structured adversarial perturbations and geometric corruptions, posing risks to their deployment in safety-critical scenarios. Existing certified defenses limit point-wise perturbations but overlook subtle geometric distortions that preserve individual points yet alter the overall structure, potentially leading to misclassification. In this work, we propose FreqCert, a novel certification framework that departs from conventional spatial domain defenses by shifting robustness analysis to the frequency domain, enabling structured certification against global l2-bounded perturbations. FreqCert first transforms the input point cloud via the graph Fourier transform (GFT), then applies structured frequency-aware subsampling to generate multiple sub-point clouds. Each sub-cloud is independently classified by a standard model, and the final prediction is obtained through majority voting, where sub-clouds are constructed based on spectral similarity rather than spatial proximity, making the partitioning more stable under l2 perturbations and better aligned with the object’s intrinsic structure. We derive a closed-form lower bound on the certified l2 robustness radius and prove its tightness under minimal and interpretable assumptions, establishing a theoretical foundation for frequency domain certification. Extensive experiments on the ModelNet40 and ScanObjectNN datasets demonstrate that FreqCert consistently achieves higher certified accuracy and empirical accuracy under strong perturbations. Our results suggest that spectral representations provide an effective pathway toward certifiable robustness in 3D point cloud recognition.

PDF Details DOI

AAAI Conference 2026 Conference Paper

SinBasis Networks: Matrix-Equivalent Feature Extraction for Wave-Like Optical Spectrograms

  • Yuzhou Zhu
  • Zheng Zhang
  • Ruyi Zhang
  • Liang Zhou

Wave-like images—from attosecond streaking spectrograms to optical spectra, audio mel-spectrograms and periodic video frames—encode critical harmonic structures that elude conventional feature extractors. We propose a unified, matrix-equivalent framework that reinterprets convolution and attention as linear transforms on flattened inputs, revealing filter weights as basis vectors spanning latent feature subspaces. To infuse spectral priors we apply elementwise sin(·) mappings to each weight matrix. Embedding these transforms into CNN, ViT and Capsule architectures yields Sin-Basis Networks with heightened sensitivity to periodic motifs and built-in invariance to spatial shifts. Experiments on a diverse collection of wave-like image datasets—including 80,000 synthetic attosecond streaking spectrograms, thousands of Raman, photoluminescence and FTIR spectra, mel-spectrograms from AudioSet and cycle-pattern frames from Kinetics—demonstrate substantial gains in reconstruction accuracy, translational robustness and zero-shot cross-domain transfer. Theoretical analysis via matrix isomorphism and Mercer-kernel truncation quantifies how sinusoidal reparametrization enriches expressivity while preserving stability in data-scarce regimes. Sin-Basis Networks thus offer a lightweight, physics-informed approach to deep learning across all wave-form imaging modalities.

PDF Details DOI

IROS Conference 2025 Conference Paper

Extreme-Hydrostatic-Pressure Resilient Dielectric Elastomer Actuator for Propeller Propulsion

  • Boyuan Du
  • Liang Zhou
  • Xuguang Dong
  • Xinge Li
  • Tong Chen
  • Tiefeng Li
  • Xin-Jun Liu
  • Huichan Zhao

Exploring high hydrostatic pressure environments such as deep sea presents significant challenges to robotic devices, for they often rely on strong yet heavy and costly protective structures to shield components from being crushed by the extreme pressure. To dismiss the need for bulky protection shells for actuation devices, we reported an extreme-hydrostatic-pressure resilient rotary dielectric elastomer actuator (DEA) for propulsion application in deep-sea pressure condition. DEAs are inherently resistant to damage caused by external pressure, due to their uniform and cavity-free structure. In this study, we analyzed the material properties of the DEA’s elastomer, evaluated the rotary actuator’s lifespan at up to 110 MPa high-pressure liquid conditions, and output performance under both ambient and 30 MPa (equivalent to 3, 000 m underwater). Our results show that the rotary actuator maintained functionality at such hydrostatic pressure, with a lifespan exceeding 300, 000 cycles and a high rotational output speed of 820 rpm. The rotary actuator was subsequently used to drive the robot with a propeller in a simulated deep-sea pressure fluidic environment, demonstrating our DEA’s performance as well as design simplicity for deep-sea applications without protection structures. While high hydrostatic pressure negatively impacted the actuator’s lifespan and slightly reduced its dynamic performance, our results confirmed that the DEA was a viable solution for deep-sea exploration, laying a solid foundation for the further development of DEA-powered devices for underwater missions.

Details

EAAI Journal 2024 Journal Article

Two-stage deep learning-based video image recognition of early fires in heritage buildings

  • Shenyuan Gao
  • Guozhong Huang
  • Xiaolong Chen
  • Huiling Jiang
  • Liang Zhou
  • Xuehong Gao

Fire poses a huge safety risk to heritage or historic buildings. Traditional approaches are unsuitable because of their high false alarm rates, delayed features, and vulnerability in heritage building environments. To reduce the serious consequences caused by fires, it is necessary to develop advanced approaches to achieve real-time early-fire detection in heritage buildings. Hence, this study proposed a two-stage deep learning-based video image recognition of real-time fires in heritage buildings. Motion detection based on a Gaussian mixture model was applied to determine moving objects in the first stage. Then, a Fire-Det model for the indoor environment and a Fire-Det Nano model for outdoor areas were developed in the second stage. To validate the proposed models, a dataset including negative images was built and tested. To illustrate the performance of this study, the proposed approaches were compared with traditional methods. It is found that the Fire-Det model can complete fire detection at the fastest reasoning speed without losing generalizability of accuracy. The proposed Fire-Det Nano model can also drastically reduce computation time at the expense of a slight reduction in accuracy. Fire-Det is the innovation in AI. Fire-Det Nano and motion detection are applications in engineering. These findings indicate that the proposed method has great potential significance in the application of early-fire detection in heritage buildings, such as the Forbidden City.

Details DOI

JBHI Journal 2022 Journal Article

ULECGNet: An Ultra-Lightweight End-to-End ECG Classification Neural Network

  • Jianbiao Xiao
  • Jiahao Liu
  • Huanqi Yang
  • Qingsong Liu
  • Ning Wang
  • Zhen Zhu
  • Yulong Chen
  • Yu Long

ECG classification is a key technology in intelligent electrocardiogram (ECG) monitoring. In the past, traditional machine learning methods such as support vector machine (SVM) and K-nearest neighbor (KNN) have been used for ECG classification, but with limited classification accuracy. Recently, the end-to-end neural network has been used for ECG classification and shows high classification accuracy. However, the end-to-end neural network has large computational complexity including a large number of parameters and operations. Although dedicated hardware such as field-programmable gate array (FPGA) and application-specific integrated circuit (ASIC) can be developed to accelerate the neural network, they result in large power consumption, large design cost, or limited flexibility. In this work, we have proposed an ultra-lightweight end-to-end ECG classification neural network that has extremely low computational complexity (∼8. 2k parameters & ∼227k multiplication/addition operations) and can be squeezed into a low-cost microcontroller (MCU) such as MSP432 while achieving 99. 1% overall classification accuracy. This outperforms the state-of-the-art ECG classification neural network. Implemented on MSP432, the proposed design consumes only 0. 4 mJ and 3. 1 mJ per heartbeat classification for normal and abnormal heartbeats respectively for real-time ECG classification.

Details DOI

ICLR Conference 2021 Conference Paper

Generalized Energy Based Models

  • Michael Arbel
  • Liang Zhou
  • Arthur Gretton

We introduce the Generalized Energy Based Model (GEBM) for generative modelling. These models combine two trained components: a base distribution (generally an implicit model), which can learn the support of data with low intrinsic dimension in a high dimensional space; and an energy function, to refine the probability mass on the learned support. Both the energy function and base jointly constitute the final model, unlike GANs, which retain only the base distribution (the "generator"). GEBMs are trained by alternating between learning the energy and the base. We show that both training stages are well-defined: the energy is learned by maximising a generalized likelihood, and the resulting energy-based loss provides informative gradients for learning the base. Samples from the posterior on the latent space of the trained model can be obtained via MCMC, thus finding regions in this space that produce better quality samples. Empirically, the GEBM samples on image-generation tasks are of much better quality than those from the learned generator alone, indicating that all else being equal, the GEBM will outperform a GAN of the same complexity. When using normalizing flows as base measures, GEBMs succeed on density modelling tasks returning comparable performance to direct maximum likelihood of the same networks.

Details
v2026.09.13