With the rapid evolution of the Industrial Internet of Things (IIoT), dynamic trust evaluation for substantial industrial devices has emerged as a critical defense line for IIoT security. Although Deep Learning (DL) models are widely adopted due to their superior performance, their inherent “black-box” nature causes severe feature entanglement, hindering the establishment of trustworthy decision-making grounds. To mitigate this opacity, Explainable AI (XAI) paradigms—primarily Post-hoc Explanation and Intrinsic Interpretability—have been introduced. However, existing approaches face a critical dilemma: post-hoc methods merely approximate model behaviors via input perturbations, often suffering from low fidelity and approximation errors in high-dimensional IIoT scenarios; traditional intrinsic models guarantee transparency but often fail to capture complex nonlinear patterns, leading to suboptimal detection accuracy. To reconcile this trade-off, we propose HITE, a Hierarchical Intrinsically Interpretable Trust Evaluation framework. Uniquely, this framework leverages the additive structure of Kolmogorov-Arnold Networks (KAN) to achieve explicit additive decomposition of raw features. By synergizing with structured decision trees, HITE constructs a hybrid reasoning chain characterized by transparent representation and explicit decision logic. It provides end-to-end traceable explanations across four dimensions: feature mapping, representation attribution, decision path, and outcome scoring. Experiments conducted on two real-world industrial benchmarks, X-IIoTID and ToN-IoT, demonstrate that HITE achieves superior evaluation performance, attaining accuracies of 99.31% and 98.96%, respectively. Crucially, HITE outperforms mainstream explanation methods in terms of faithfulness and stability, realizing a critical paradigm shift from “Numerical Prediction” to “Trusted Reasoning”.

Human-annotated attributes serve as effective semantic label embeddings for zero-shot learning (ZSL); however, their annotation is labor-intensive and difficult to scale. Recent studies have explored weakly supervised semantic label embeddings to reduce human effort, but these methods often fail to capture visual similarity and underperform compared to human-annotated semantics. In this work, we propose a minimally supervised yet effective approach: GPT- and CLIP-powered attributes (GCAtt). Specifically, we introduce a three-step interaction process with ChatGPT-comprising preliminary design, hierarchical refinement, and specific value determination-to generate attributes that are both category-shared and discriminative for classification. Additionally, we develop a method that encodes attributes and their values as potential text pairings, leveraging CLIP's retrieval capabilities for annotation. Experimental results on four widely used benchmarks demonstrate that GCAtt consistently outperforms human-annotated semantics. Code and data are available at https://github.com/RowenaHe/GCAtt.

Fangwei Zhu, Shangqing Tu, Jiaxin Shi, Juanzi Li, Lei Hou, Tong Cui. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Traditional imaging methods and computer vision algorithms are often ineffective when images are acquired in scattering media, such as underwater, fog, and biological tissue. Here, we explore the use of light field imaging and algorithms for image restoration and depth estimation that address the image degradation from the medium. Towards this end, we make the following three contributions. First, we present a new single image restoration algorithm which removes backscatter and attenuation from images better than existing methods do, and apply it to each view in the light field. Second, we combine a novel transmission based depth cue with existing correspondence and defocus cues to improve light field depth estimation. In densely scattering media, our transmission depth cue is critical for depth estimation since the images have low signal to noise ratios which significantly degrades the performance of the correspondence and defocus cues. Finally, we propose shearing and refocusing multiple views of the light field to recover a single image of higher quality than what is possible from a single view. We demonstrate the benefits of our method through extensive experimental results in a water tank.

Image dehazing is a useful technique which can eliminate the bad effect of haze on images and enhance the performances of image/video processing algorithms in the hazy weather. In this study, a single image dehazing method is proposed. The authors estimate the initial transmission properly based on latent region‐segmentation and refine the estimated initial transmission by an objective function with a novel weighted L 1 ‐norm regularisation term. The half‐quadratic splitting minimisation method is employed to solve this optimisation problem. They also define an evaluation function to estimate the reliable global atmospheric light. With the refined transmission map and atmospheric light they recover the haze‐free image by the haze imaging model. The authors’ method is compared with three state‐of‐the‐art methods and is also validated by two image quality assessment methods. The comparative experimental results and evaluations demonstrate that their method can recover comparable and even better results with clear details, low contrast loss and high contrast in most cases.

Auto-encoder is a popular representation learning technique which can capture the generative model of data via a encoding and decoding procedure typically driven by reconstruction errors in an unsupervised way. In this paper, we propose a semi-supervised manifold learning based auto-encoder (named semAE). semAE is based on a regularized auto-encoder framework which leverages semi-supervised manifold learning to impose regularization based on the encoded representation. Our proposed approach suits more practical scenarios in which a small number of labeled data are available in addition to a large number of unlabeled data. Experiments are conducted on several well-known benchmarking datasets to validate the efficacy of semAE from the aspects of both representation and classification. The comparisons to state-of-the-art representation learning methods on classification performance in semi-supervised settings demonstrate the superiority of our approach.

LDAP Multi-Master Technique is a replication approach using Syncrepl to replicate data to multiple servers. As the specialized database optimized for read access, the directory is used to represent heterogeneous entities in directory information tree (DIT). Compared with existing Sync replication models, the directory template replication model based on query templates proposed in this paper only replicates the entries matching the template specification. Replicated templates can reduce the complexity of the query containment problem and generalize the user queries corresponding to semantic regions demonstrating locality of reference. The synchronization protocol, which uses standard means of extending LDAP, is used to support (N-way) Multi-Master server replication modes. In this case, masters can be located in several physical sites, and other masters will continue the database if one master fails. Finally, we demonstrate the advantages of the novel model considering higher hit rates and the large update traffic with an experimental study based on real data and a prototype implementation of the LDAP directory cache.

In this paper, a new method to model the global deformation between a rigid object and an elastic object with a hole is presented. This method extends the idea of beam-skeletons [10] by introducing curved cantilever beams for efficient modeling of global deformation of elastic objects with holes. The method is implemented and tested on different examples. Results from three examples are given to demonstrate the efficiency and effectiveness of the approach.

This paper addresses a largely open problem in haptic simulation and rendering: contact force and deformation modeling for haptic simulation of grasping a deformable object with a realistic virtual human hand, especially in power grasps. The virtual hand model consists of meshes of realistic shapes for the finger links and palm of a hand. We tackle the problem by adopting the non-linear contact force model and the beam-skeleton model for global shape deformation introduced in [5]. The results verify the efficiency of contact force and deformation modeling for both power grasp and precision grasp of deformable objects with reasonable realism.