Institution profile

Nanjing University of Information Science and Technology

Academic institutionasia · cn
Official website
Research library262linked papers
Opportunities0open roles
Selected work

Representative Papers

Artificial Protozoa Optimizer (APO): A novel bio-inspired metaheuristic algorithm for engineering optimization

Apr 01, 2024Knowledge-Based Systems

This paper addresses high-dimensional nonlinear engineering optimization problems by proposing a novel bio-inspired metaheuristic: the Artificial Protozoan Optimizer (APO). APO systematically models multifaceted biological mechanisms of protozoa—including chemotactic foraging, dynamic fission, and adaptive phagocytosis—to establish a balanced optimization framework with strong global exploration and local exploitation capabilities. It introduces a dynamic population fission strategy and an adaptive phagocytosis operator to enable parameter self-regulation and synergistic stochastic search. Evaluated on the CEC2020 benchmark suite and multiple constrained engineering design problems, APO significantly outperforms mainstream algorithms such as PSO, GWO, and HHO—achieving a 32% improvement in convergence speed and enhancing optimal solution accuracy by one to two orders of magnitude. Furthermore, APO demonstrates practical efficacy in multilevel image segmentation tasks.

57 citations2 influentialRead paper

Diff-PC: Identity-preserving and 3D-aware controllable diffusion for zero-shot portrait customization

Dec 01, 2024Information Fusion

Existing portrait customization methods often struggle to simultaneously preserve identity fidelity and achieve precise facial control. To address this limitation, this work proposes Diff-PC, a novel framework that leverages a 3D face-guided identity encoder and a feature injection mechanism to enable high-fidelity, fine-grained controllable portrait generation under zero-shot conditions. The approach effectively integrates 3D-aware priors with identity features and is compatible with diverse backgrounds and multi-style base models. Trained with a dedicated identity-centric dataset and enhanced by an ID-Encoder, an ID-Ctrl alignment module, and an ID-Injector refinement module, Diff-PC consistently outperforms state-of-the-art methods in terms of identity preservation, facial controllability, and text-to-image alignment.

6 citationsRead paper

A Pairwise Comparison Relation-assisted Multi-objective Evolutionary Neural Architecture Search Method with Multi-population Mechanism

Jul 22, 2024arXiv.org

Neural architecture search (NAS) suffers from high evaluation overhead and model redundancy due to single-objective optimization (e.g., accuracy only). To address this, we propose an efficient multi-objective NAS framework. Our method introduces: (1) a novel lightweight surrogate model based on pairwise comparison that predicts relative architectural rankings instead of absolute accuracy—substantially reducing evaluation cost; and (2) a master–auxiliary dual-population co-evolutionary mechanism that enhances population diversity while ensuring convergence. Evaluated on CIFAR-10/100 and ImageNet, our approach completes search in just 0.17 GPU-days on a single GPU. On ImageNet, it discovers a compact architecture achieving 78.91% Top-1 accuracy with only 570M MAdds. Compared to state-of-the-art methods, our framework improves search efficiency by multiple orders of magnitude and significantly strengthens multi-objective optimization across accuracy, parameter count, and computational cost.

2 citationsRead paper

Exposing and Defending the Achilles'Heel of Video Mixture-of-Experts

Feb 01, 2026

This work addresses the insufficient robustness of Video Mixture-of-Experts (MoE) models under adversarial attacks, a vulnerability exacerbated by the neglect of both individual and collaborative weaknesses in router and expert modules. To this end, we propose Temporal Lipschitz-Guided Attack (TLGA) and its joint variant (J-TLGA), which respectively target the router and jointly perturb both router and experts, thereby systematically uncovering component-level vulnerabilities in MoE architectures for the first time. Building on these insights, we design a plug-and-play Joint Temporal Lipschitz Adversarial Training (J-TLAT) framework with low inference overhead. Extensive experiments demonstrate that our approach significantly enhances adversarial robustness across multiple video datasets and MoE architectures, while reducing inference costs by over 60% compared to dense models, effectively mitigating both individual and collaborative vulnerabilities.

1 citationsRead paper

From Spurious to Causal: Low-rank Orthogonal Subspace Intervention for Generalizable Face Forgery Detection

Jan 17, 2026

This work addresses the limited generalization of face forgery detection models caused by spurious correlations induced by unobservable confounding factors in their representations. The authors propose a novel representation-space intervention paradigm that, for the first time, unifies diverse spurious correlations into a low-rank subspace and removes them from the original features via orthogonal low-rank projection. The model instead learns from the orthogonal complement subspace to capture genuine forgery cues. This approach obviates the need to explicitly model each confounding factor individually and introduces only 0.43 million trainable parameters. Despite its minimal overhead, it achieves state-of-the-art performance across multiple benchmarks, significantly enhancing both robustness and generalization capability.

1 citationsRead paper
Recent publications

Latest Papers