Qun Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Qun Liu has a total of 125 co-authors. Here are the top co-authors by number of shared publications:
- Guoyin Wang
Chongqing Normal University
26 shared publications Guoyin Wang
13 shared publications- Li Liu
Chongqing University of Posts and Telecommunications
8 shared publications - Ye Wang
Chongqing University of Posts and Telecommunications
6 shared publications - Hao Zhou
College of Computer Science, Chongqing University, Chongqing, China
5 shared publications Xu Gong
Chongqing Key Laboratory of Toxicology and Drug Analysis, Chongqing Police College, Chongqing 401331, China
5 shared publications
Latest Publications
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 38 (10) , 6572-6585
Enhancing Graph Neural Network Explainers Using a Distribution Shift Consistency-Guided Generator
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 38 (10) , 6572-6585
Enhancing Graph Neural Network Explainers Using a Distribution Shift Consistency-Guided Generator
- Open Access
- 7 September 2026
- Published by Scilight Press Pty Ltd in Journal of Machine Learning and Information Security
- Vol. 2 (3) , 18
MG-CMIF: A Multi-Granularity Enhanced Cross-Modal Information Fusion Framework for Molecular Property Prediction
Molecular property prediction provides an important computational basis for compound screening and drug development by estimating physicochemical characteristics and biological activities from molecular structures. Although deep learning has improved molecular modeling, existing methods often describe molecules through a limited structural view or combine multiple views without sufficiently exploiting their complementary relationships. In addition, graph-based approaches commonly concentrate on local atomic connectivity, making it difficult to represent chemically meaningful structural units that may strongly influence molecular properties. This paper develops MG-CMIF, a multi-granularity cross-modal framework for molecular property prediction. The proposed model describes each molecule from symbolic, topological, and spatial perspectives and learns an integrated representation through three key designs. First, hierarchical graph modeling combines detailed atomic interactions with substructure-level chemical patterns to enrich topology-oriented features. Second, interaction across molecular views enables information relevant to property prediction to be exchanged selectively rather than merged through shallow operations. Third, alignment-oriented training objectives encourage representations derived from the same molecule to preserve compatible chemical semantics during fusion. Experiments on multiple public benchmark datasets show that MG-CMIF achieves better prediction results than competitive methods in both classification and regression settings. Further ablation analyses confirm that hierarchical structural modeling and cross-view integration both contribute to the effectiveness of the proposed framework.
- 7 September 2026
- Published by Scilight Press Pty Ltd in Journal of Machine Learning and Information Security
- Vol. 2 (3) , 18
Open AccessMG-CMIF: A Multi-Granularity Enhanced Cross-Modal Information Fusion Framework for Molecular Property Prediction
Molecular property prediction provides an important computational basis for compound screening and drug development by estimating physicochemical characteristics and biological activities from molecular structures. Although deep learning has improved molecular modeling, existing methods often describe molecules through a limited structural view or combine multiple views without sufficiently exploiting their complementary relationships. In addition, graph-based approaches commonly concentrate on local atomic connectivity, making it difficult to represent chemically meaningful structural units that may strongly influence molecular properties. This paper develops MG-CMIF, a multi-granularity cross-modal framework for molecular property prediction. The proposed model describes each molecule from symbolic, topological, and spatial perspectives and learns an integrated representation through three key designs. First, hierarchical graph modeling combines detailed atomic interactions with substructure-level chemical patterns to enrich topology-oriented features. Second, interaction across molecular views enables information relevant to property prediction to be exchanged selectively rather than merged through shallow operations. Third, alignment-oriented training objectives encourage representations derived from the same molecule to preserve compatible chemical semantics during fusion. Experiments on multiple public benchmark datasets show that MG-CMIF achieves better prediction results than competitive methods in both classification and regression settings. Further ablation analyses confirm that hierarchical structural modeling and cross-view integration both contribute to the effectiveness of the proposed framework.
- 28 August 2026
- Published by Springer Nature in Applied Intelligence
- Vol. 56 (13) , 395
MD-VAE: Concepts embedded variational autoencoder with multiple decoders
- 28 August 2026
- Published by Springer Nature in Applied Intelligence
- Vol. 56 (13) , 395
MD-VAE: Concepts embedded variational autoencoder with multiple decoders
- 10 August 2026
- Published by Elsevier in Applied Soft Computing
- Vol. 203
NOMG: Nonlinear oversampling based on multi-granularity optimization
- 10 August 2026
- Published by Elsevier in Applied Soft Computing
- Vol. 203
NOMG: Nonlinear oversampling based on multi-granularity optimization
- 26 July 2026
- Published by Springer Nature
- p. 161-175
A Multi-granularity Collaborative Genetic Algorithm for Truck-Drone Routing Problem
- 26 July 2026
- Published by Springer Nature
- p. 161-175
A Multi-granularity Collaborative Genetic Algorithm for Truck-Drone Routing Problem
- Open Access
- 21 July 2026
- Published by Elsevier in SSRN Electronic Journal
OURF: Optimization-driven undersampling robust framework for multi-class imbalanced noisy learning
Multi-class imbalanced noisy learning is fundamentally challenged by the intricate interplay among class imbalance, structural degradation, and label noise. Existing undersampling methods predominantly rely on fixed-ratio reduction or heuristic sample removal, often leading to information collapse, manifold distortion, and deteriorated decision boundaries while remaining vulnerable to noise propagation. To address these limitations without generating additional minority samples, this paper proposes an optimization-driven undersampling robust framework (OURF) for multi-class imbalanced noisy learning. Unlike conventional undersampling that focus primarily on numerical class balancing, OURF reinterprets undersampling as a structural-risk control problem and, for the first time, establishes a two-stage optimization framework from the perspective of optimal balance, enabling globally coordinated optimization among class balancing, structure preservation, and noise suppression. Specifically, a structure-risk balanced undersampling quantity optimization (SRB-UQO) model is developed to adaptively determine class-wise retention quantities through a structure-noise adaptive target prior, thereby preventing information collapse caused by excessive compression. Building upon the learned quantity constraints, a noise-robust manifold-preserving sample–feature optimization (NRMP-SFO) model further performs structure-robust representation refinement, producing retained subsets that preserve geometric consistency, topological integrity, and discriminative stability. From a theoretical perspective, we establish a unified structural-risk control framework for optimization-driven undersampling. The analysis reveals that the fundamental limitation of conventional undersampling stems from its inability to jointly control coupled structural risks, including information collapse, manifold fragmentation, and noise exposure. Furthermore, we prove that OURF tightens a provable generalization-risk upper bound through the joint optimization of multiple computable structural-risk surrogates. Extensive experiments demonstrate that OURF serves as a general optimization framework that not only consistently enhances representative undersampling methods but also yields a competitive standalone undersampling scheme. Across diverse noise scenarios, OURF exhibits superior structural robustness, stability, and cross-noise generalization capability compared with SOTA methods.
- 21 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessOURF: Optimization-driven undersampling robust framework for multi-class imbalanced noisy learning
Multi-class imbalanced noisy learning is fundamentally challenged by the intricate interplay among class imbalance, structural degradation, and label noise. Existing undersampling methods predominantly rely on fixed-ratio reduction or heuristic sample removal, often leading to information collapse, manifold distortion, and deteriorated decision boundaries while remaining vulnerable to noise propagation. To address these limitations without generating additional minority samples, this paper proposes an optimization-driven undersampling robust framework (OURF) for multi-class imbalanced noisy learning. Unlike conventional undersampling that focus primarily on numerical class balancing, OURF reinterprets undersampling as a structural-risk control problem and, for the first time, establishes a two-stage optimization framework from the perspective of optimal balance, enabling globally coordinated optimization among class balancing, structure preservation, and noise suppression. Specifically, a structure-risk balanced undersampling quantity optimization (SRB-UQO) model is developed to adaptively determine class-wise retention quantities through a structure-noise adaptive target prior, thereby preventing information collapse caused by excessive compression. Building upon the learned quantity constraints, a noise-robust manifold-preserving sample–feature optimization (NRMP-SFO) model further performs structure-robust representation refinement, producing retained subsets that preserve geometric consistency, topological integrity, and discriminative stability. From a theoretical perspective, we establish a unified structural-risk control framework for optimization-driven undersampling. The analysis reveals that the fundamental limitation of conventional undersampling stems from its inability to jointly control coupled structural risks, including information collapse, manifold fragmentation, and noise exposure. Furthermore, we prove that OURF tightens a provable generalization-risk upper bound through the joint optimization of multiple computable structural-risk surrogates. Extensive experiments demonstrate that OURF serves as a general optimization framework that not only consistently enhances representative undersampling methods but also yields a competitive standalone undersampling scheme. Across diverse noise scenarios, OURF exhibits superior structural robustness, stability, and cross-noise generalization capability compared with SOTA methods.
- 2 July 2026
- Published by Springer Nature
- p. 341-353
Brain-Inspired Fractional Convolution for Pathological Image Classification
- 2 July 2026
- Published by Springer Nature
- p. 341-353
Brain-Inspired Fractional Convolution for Pathological Image Classification
- 16 February 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 314
IFSA-CE: Interpretable fine-grained sentiment analysis with concept embedding
- 16 February 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 314
IFSA-CE: Interpretable fine-grained sentiment analysis with concept embedding
- 20 November 2025
- Published by Elsevier in Neurocomputing
- Vol. 664
MGFD: A multi-granularity pathological auxiliary diagnosis model based on Grünwald–Letnikov fractional derivative
- 20 November 2025
- Published by Elsevier in Neurocomputing
- Vol. 664
MGFD: A multi-granularity pathological auxiliary diagnosis model based on Grünwald–Letnikov fractional derivative
- 24 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36 (3) , 2842-2853
Visual Evidence-Aware for Object Hallucinations Rectification in LLM-Based Video Captioning
Recent neural models for video captioning are typically built using a framework that combines a pre-trained visual encoder with a large language model(LLM) decoder. However, large language models in video captioning often generate non-existent entities, known as object hallucinations, which severely limit performance. To mitigate object hallucinations, two key issues remain: 1. Biased training data and Knowledge bias in LLM leads models to generate hallucinations; 2. Current methods focus on removal rather than restoring the correct visual content, reducing caption completeness. To address these issues, we propose a visual evidence-aware for object hallucination rectification in LLM-based video captioning. Generally, our model aims to diagnose and correct those generated object hallucinations, and then supplement missing visual content by constraining the process of text description generation. Specifically, we first generate captions by words based on the input video. When decoding each object description, the decoder utilizes visual features for hallucination diagnosis and correction, proposing visual evidence to modify hallucinatory descriptions. This process ensures the generated captions align with the visual content, alleviating the generation of object hallucinations. Compared with the baseline models, our method performs state-of-the-art performance in video captioning, especially avoiding neglecting objects in the visual content caused by the generated hallucinatory descriptions.
- 24 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36 (3) , 2842-2853
Visual Evidence-Aware for Object Hallucinations Rectification in LLM-Based Video Captioning
Recent neural models for video captioning are typically built using a framework that combines a pre-trained visual encoder with a large language model(LLM) decoder. However, large language models in video captioning often generate non-existent entities, known as object hallucinations, which severely limit performance. To mitigate object hallucinations, two key issues remain: 1. Biased training data and Knowledge bias in LLM leads models to generate hallucinations; 2. Current methods focus on removal rather than restoring the correct visual content, reducing caption completeness. To address these issues, we propose a visual evidence-aware for object hallucination rectification in LLM-based video captioning. Generally, our model aims to diagnose and correct those generated object hallucinations, and then supplement missing visual content by constraining the process of text description generation. Specifically, we first generate captions by words based on the input video. When decoding each object description, the decoder utilizes visual features for hallucination diagnosis and correction, proposing visual evidence to modify hallucinatory descriptions. This process ensures the generated captions align with the visual content, alleviating the generation of object hallucinations. Compared with the baseline models, our method performs state-of-the-art performance in video captioning, especially avoiding neglecting objects in the visual content caused by the generated hallucinatory descriptions.