arXiv subjects

Zhengxin Zeng

Publications and source records attributed to Zhengxin Zeng.

6 recordsLinked to original sources

arXiv2026

Unleash LLMs Potential for Sequential Recommendation by Coordinating Dual Dynamic Index Mechanism

Owing to the unprecedented capability in semantic understanding and logical reasoning, large language models (LLMs) have shown fantastic potential in developing next-generation sequential recommender systems (RSs). However, existing LLM-based sequential RSs mostly separate index generation from sequential recommendation, leading to insufficient integration between semantic information and collaborative information. On the other hand, the neglect of user-related information hinders LLM-based sequential RSs from exploiting high-order user-item interaction patterns. In this paper, we propose the End-to-End Dual Dynamic (ED$^2$) recommender, the first LLM-based sequential RS which adopts dual dynamic index mechanism, targeting resolving the above limitations simultaneously. The dual dynamic index mechanism can not only assembly index generation and sequential recommendation into a unified LLM-backbone pipeline, but also make it practical for LLM-based sequential recommender to take advantage of user-related information. Specifically, to facilitate the LLM comprehension ability to dual dynamic index, we propose a multigrained token regulator which constructs alignment supervision based on LLMs semantic knowledge across multiple representation granularities. Moreover, the associated user collection data and a series of novel instruction tuning tasks are specially customized to capture the high-order user-item interaction patterns. Extensive experiments on three public datasets demonstrate the superiority of ED$^2$, achieving an average improvement of 19.62% in Hit-Rate and 21.11% in NDCG.

cs.IR↗

arXiv2025

When Graph meets Multimodal: Benchmarking and Meditating on Multimodal Attributed Graphs Learning

Multimodal Attributed Graphs (MAGs) are ubiquitous in real-world applications, encompassing extensive knowledge through multimodal attributes attached to nodes (e.g., texts and images) and topological structure representing node interactions. Despite its potential to advance diverse research fields like social networks and e-commerce, MAG representation learning (MAGRL) remains underexplored due to the lack of standardized datasets and evaluation frameworks. In this paper, we first propose MAGB, a comprehensive MAG benchmark dataset, featuring curated graphs from various domains with both textual and visual attributes. Based on MAGB dataset, we further systematically evaluate two mainstream MAGRL paradigms: $\textit{GNN-as-Predictor}$, which integrates multimodal attributes via Graph Neural Networks (GNNs), and $\textit{VLM-as-Predictor}$, which harnesses Vision Language Models (VLMs) for zero-shot reasoning. Extensive experiments on MAGB reveal following critical insights: $\textit{(i)}$ Modality significances fluctuate drastically with specific domain characteristics. $\textit{(ii)}$ Multimodal embeddings can elevate the performance ceiling of GNNs. However, intrinsic biases among modalities may impede effective training, particularly in low-data scenarios. $\textit{(iii)}$ VLMs are highly effective at generating multimodal embeddings that alleviate the imbalance between textual and visual attributes. These discoveries, which illuminate the synergy between multimodal attributes and graph topologies, contribute to reliable benchmarks, paving the way for future MAG research. The MAGB dataset and evaluation pipeline are publicly available at https://github.com/sktsherlock/MAGB.

cs.LG↗

arXiv2020

Automatic Arm Motion Recognition Based on Radar Micro-Doppler Signature Envelopes

In considering human-machine interface (HMI) for smart environment, a simple but effective method is proposed for automatic arm motion recognition with a Doppler radar sensor. Arms, in lieu of hands, have stronger radar cross-section and can be recognized from relatively longer distances. An energy-based thresholding algorithm is applied to the spectrograms to extract the micro-Doppler (MD) signature envelopes. The positive and negative frequency envelopes are concatenated to form a feature vector. The nearest neighbor (NN) classifier with Manhattan distance (L1) is then used to recognize the arm motions. It is shown that this simple method yields classification accuracy above 97 percent for six classes of arm motions. Despite its simplicity, the proposed method is superior to those of handcrafted featurebased classifications and low-dimension representation techniques based on principal component analysis (PCA), and is comparable to convolutional neural network (CNN).

eess.SP↗

arXiv2020

Split to Be Slim: An Overlooked Redundancy in Vanilla Convolution

Many effective solutions have been proposed to reduce the redundancy of models for inference acceleration. Nevertheless, common approaches mostly focus on eliminating less important filters or constructing efficient operations, while ignoring the pattern redundancy in feature maps. We reveal that many feature maps within a layer share similar but not identical patterns. However, it is difficult to identify if features with similar patterns are redundant or contain essential details. Therefore, instead of directly removing uncertain redundant features, we propose a \textbf{sp}lit based \textbf{conv}olutional operation, namely SPConv, to tolerate features with similar patterns but require less computation. Specifically, we split input feature maps into the representative part and the uncertain redundant part, where intrinsic information is extracted from the representative part through relatively heavy computation while tiny hidden details in the uncertain redundant part are processed with some light-weight operation. To recalibrate and fuse these two groups of processed features, we propose a parameters-free feature fusion module. Moreover, our SPConv is formulated to replace the vanilla convolution in a plug-and-play way. Without any bells and whistles, experimental results on benchmarks demonstrate SPConv-equipped networks consistently outperform state-of-the-art baselines in both accuracy and inference time on GPU, with FLOPs and parameters dropped sharply.

cs.CV↗

arXiv2019

Arm Motion Classification Using Curve Matching of Maximum Instantaneous Doppler Frequency Signatures

Hand and arm gesture recognition using the radio frequency (RF) sensing modality proves valuable in manmachine interface and smart environment. In this paper, we use curve matching techniques for measuring the similarity of the maximum instantaneous Doppler frequencies corresponding to different arm gestures. In particular, we apply both Frechet and dynamic time warping (DTW) distances that, unlike the Euclidean (L2) and Manhattan (L1) distances, take into account both the location and the order of the points for rendering two curves similar or dissimilar. It is shown that improved arm gesture classification can be achieved by using the DTW method, in lieu of L2 and L1 distances, under the nearest neighbor (NN) classifier.

eess.SP↗

arXiv2019

Hand Gesture Recognition based on Radar Micro-Doppler Signature Envelopes

We introduce a simple but effective technique in automatic hand gesture recognition using radar. The proposed technique classifies hand gestures based on the envelopes of their micro-Doppler signatures. These envelopes capture the distinctions among different hand movements and their corresponding positive and negative Doppler frequencies which are generated during each gesture act. We detect the positive and negative envelopes separately, and form a feature vector of their augmentation. We use the $k$-nearest neighbor ($k$NN) classifier and Manhattan distance (L1) measure, in lieu of Euclidean distance (L2), so as not to diminish small but critical envelope values. It is shown that this method outperforms both low-dimension representation techniques based on principal component analysis (PCA) and sparse reconstruction using Gaussian-windowed Fourier dictionary, and can achieve very high classification rates.

eess.SP↗