Author name cluster

Liusheng Huang

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.

7 papers

2 author rows

AAAI Conference 2025 Conference Paper

Stop Diverse OOD Attacks: Knowledge Ensemble for Reliable Defense

Zhenbo Shi
Xiaoman Liu
Yuxuan Zhang
Shuchang Wang
Rui Shu
Zhidong Yu
Wei Yang
Liusheng Huang

Enhancing defense through model ensemble is an emerging trend, where the challenge lies in how to use ensemble knowledge to counter Out-of-Distribution (OOD) attacks. In this paper, we propose the Reliable Defense Ensemble (REE) to address this issue. REE optimizes the ensemble knowledge of models through aggregation and enhances multidimensional robust performance through collaboration. It employs the Dynamic Synergy Amplification for weight allocation and strategy adjustment. Furthermore, we design a new Kernel Anomaly Smoothing Detection Module, which detects anomalous attacks using a smoothing feature function based on Gaussian kernel mean embedding and a multi-layer feedback structure. Particularly, we build a framework that uses reinforcement learning to iteratively fine-tune the parameters of inter-model communication and consensus. Extensive experimental results show that REE outperforms current state-of-the-art methods by a large margin in defending against OOD attacks.

PDF Details DOI

I&C Journal 2024 Journal Article

Characterization of exact two-query quantum algorithms

Shaoliang Ye
Wei Yang
Liusheng Huang

Quantum query model is a crucial model for quantum computing, where one query to some input variable of a Boolean function f defined on { 0, 1 } n returns the variable value. The exact query complexity, denoted as Q E ( f ), is defined to be the minimum number of queries required to determine the function value. An important problem in this area is to give a succinct characterization of a k-query exact quantum algorithm for an arbitrary k. To date, the cases k = 1 and k = n are already solved and the case k = 2 remains unknown. Our result is that there are 27 nondegenerate Boolean functions up to isomorphism with Q E ( f ) being two, among which only two functions can be solved by a 2-query classical algorithm. The input bit number n of the above 27 functions ranges from 2 to 6, where the case n ≤ 3 is already proved and the case n = 4 is already found by numerically solving semidefinite programming, which is a complete characterization of quantum query algorithm. Assuming the correctness of the numerical result for n = 4, we prove that there are four functions in the case n = 5, one in the case n = 6 and none in the case n ≥ 7. We further show that the 25 functions for which quantum algorithm has advantage over classical algorithm contain essentially only four different structures.

Details DOI

AAAI Conference 2022 Conference Paper

Shape Prior Guided Attack: Sparser Perturbations on 3D Point Clouds

Zhenbo Shi
Zhi Chen
Zhenbo Xu
Wei Yang
Zhidong Yu
Liusheng Huang

Deep neural networks are extremely vulnerable to malicious input data. As 3D data is increasingly used in vision tasks such as robots, autonomous driving and drones, the internal robustness of the classification models for 3D point cloud has received widespread attention. In this paper, we propose a novel method named SPGA (Shape Prior Guided Attack) to generate adversarial point cloud examples. We use shape prior information to make perturbations sparser and thus achieve imperceptible attacks. In particular, we propose a Spatially Logical Block (SLB) to apply adversarial points through sliding in the oriented bounding box. Moreover, we design an algorithm called FOFA for this type of task, which further refines the adversarial attack in the process of breaking down complicated problems into sub-problems. Compared with the methods of global perturbation, our attack method consumes significantly fewer computations, making it more efficient. Most importantly of all, SPGA can generate examples with a higher attack success rate (even in a defensive situation), less perturbation budget and stronger transferability.

PDF Details

ICRA Conference 2021 Conference Paper

MDANet: Multi-Modal Deep Aggregation Network for Depth Completion

Yanjie Ke
Kun Li
Wei Yang 0011
Zhenbo Xu
Dayang Hao
Liusheng Huang
Gang Wang

Depth completion aims to recover the dense depth map from sparse depth data and RGB image respectively. However, due to the huge difference between the multi-modal signal input, vanilla convolutional neural network and simple fusion strategy cannot extract features from sparse data and aggregate multi-modal information effectively. To tackle this problem, we design a novel network architecture that takes full advantage of multi-modal features for depth completion. An effective Pre-completion algorithm is first put forward to increase the density of the input depth map and to provide distribution priors. Moreover, to effectively fuse the image features and the depth features, we propose a multi-modal deep aggregation block that consists of multiple connection and aggregation pathways for deeper fusion. Furthermore, based on the intuition that semantic image features are beneficial for accurate contour, we introduce the deformable guided fusion layer to guide the generation of the dense depth map. The resulting architecture, called MDANet, outperforms all the stateof-the-art methods on the popular KITTI Depth Completion Benchmark, meanwhile with fewer parameters than recent methods. The code of this work will be available at https://github.com/USTC-Keyanjie/MDANet_ICRA2021.

Details

TCS Journal 2020 Journal Article

Quality-aware online task assignment mechanisms using latent topic model

Yang Du
Yu-E Sun
He Huang
Liusheng Huang
Hongli Xu
Xiaocan Wu

Crowdsourcing has been proven to be a useful tool for the tasks which are hard for computers. Unfortunately, workers with uneven expertise are likely to provide low-quality or even deliberately wrong data. A reliability model that precisely describes workers' performance on the tasks can benefit the development of both task assignment mechanism and truth discovery method. However, existing methods cannot model workers' fine-grained reliability levels accurately. In this paper, we consider dividing tasks into clusters (i. e. , topics) based on workers' behaviors and propose a novel latent topic model to describe the topic structure and workers' topical-level expertise. Then, we develop two online task assignment mechanisms that dynamically assign each incoming worker a set of tasks where he can achieve the Maximum Expected Gain (MEG) or Maximum Expected and Potential Gain (MEPG). The experimental results demonstrate that our methods can significantly decrease the number of task assignments and achieve higher accuracy and macro-averaging F1-score than the state-of-the-art approaches.

Details DOI

AAAI Conference 2020 Conference Paper

ZoomNet: Part-Aware Adaptive Zooming Neural Network for 3D Object Detection

Zhenbo Xu
Wei Zhang
Xiaoqing Ye
Xiao Tan
Wei Yang
Shilei Wen
Errui Ding
Ajin Meng

3D object detection is an essential task in autonomous driving and robotics. Though great progress has been made, challenges remain in estimating 3D pose for distant and occluded objects. In this paper, we present a novel framework named ZoomNet for stereo imagery-based 3D detection. The pipeline of ZoomNet begins with an ordinary 2D object detection model which is used to obtain pairs of leftright bounding boxes. To further exploit the abundant texture cues in rgb images for more accurate disparity estimation, we introduce a conceptually straight-forward module – adaptive zooming, which simultaneously resizes 2D instance bounding boxes to a uniﬁed resolution and adjusts the camera intrinsic parameters accordingly. In this way, we are able to estimate higher-quality disparity maps from the resized box images then construct dense point clouds for both nearby and distant objects. Moreover, we introduce to learn part locations as complementary features to improve the resistance against occlusion and put forward the 3D ﬁtting score to better estimate the 3D detection quality. Extensive experiments on the popular KITTI 3D detection dataset indicate ZoomNet surpasses all previous state-of-the-art methods by large margins (improved by 9. 4% on APbv (IoU=0. 7) over pseudo-LiDAR). Ablation study also demonstrates that our adaptive zooming strategy brings an improvement of over 10% on AP3d (IoU=0. 7). In addition, since the ofﬁcial KITTI benchmark lacks ﬁne-grained annotations like pixel-wise part locations, we also present our KFG dataset by augmenting KITTI with detailed instance-wise annotations including pixel-wise part location, pixel-wise disparity, etc. . Both the KFG dataset and our codes will be publicly available at https: //github. com/detectRecog/ZoomNet.

PDF Details

TCS Journal 2012 Journal Article

Degree-bounded minimum spanning tree for unit disk graph

Hongli Xu
Liusheng Huang
Wang Liu
Yindong Zhang
Yanjing Sun

Degree-bounded minimum spanning tree (DBMST) has been widely used in many applications of wireless sensor networks, such as data aggregation, topology control, etc. However, before construction of DBMST, it is NP-hard to determine whether or not there is a degree- k spanning tree for an arbitrary graph, where k is 3 or 4. The wireless sensor network is usually modeled by a unit disk graph (UDG), where two vertices are connected in UDG G ( R ) if their Euclidean distance is not more than a given constant R in the field. The previous works have predicated the necessary conditions for the existence of DBMST on UDG. Given that sub-graphs G ( R / 2 ) and G ( R / 3 ) can keep connected, there exist degree-3 or degree-4 spanning trees for UDG G ( R ). In this paper, we design two algorithms to construct the degree-3 and degree-4 spanning trees for UDG respectively. The more relaxed conditions are explored for the existence of DBMST for unit disk graphs according to the proposed algorithms. That is, given that sub-graphs G ( R / 1. 81 ) and G ( R / 2 ) keep connected, the existence of degree-3 and degree-4 spanning trees is guaranteed for UDG G ( R ). The theoretical analyses show that the performances of constructed degree-3 and degree-4 spanning trees are at most ( 4 + 6 α ) / 4 and ( 1 + 2 α ) / 2 times as that of minimum spanning tree (MST) respectively, where α ≥ 2 is a constant. The simulation results show the high efficiency of two proposed algorithms. For example, total link weights of degree-3 and degree 4 spanning trees are about 1. 05 and 1. 01 times as that of MST where α is 2.

Details DOI