Test-time adaptation (TTA) aims to fine-tune a trained model online using unlabeled testing data to adapt to new environments or out-of-distribution data, demonstrating broad application potential in real-world scenarios. However, in this optimization process, unsupervised learning objectives like entropy minimization frequently encounter noisy learning signals. These signals produce unreliable gradients, which hinder the model’s ability to converge to an optimal solution quickly and introduce significant instability into the optimization process. In this paper, we seek to resolve these issues from the perspective of optimizer design. Unlike prior TTA using manually designed optimizers like SGD, we employ a learning-to-optimize approach to automatically learn an optimizer, called Meta Gradient Generator (MGG). Specifically, we aim for MGG to effectively utilize historical gradient information during the online optimization process to optimize the current model. To this end, in MGG, we design a lightweight and efficient sequence modeling layer -- gradient memory layer. It exploits a self-supervised reconstruction loss to compress historical gradient information into network parameters, thereby enabling better memorization ability over a long-term adaptation process. We only need a small number of unlabeled samples to pre-train MGG, and then the trained MGG can be deployed to process unseen samples. Promising results on ImageNet-C/R/Sketch/A indicate that our method surpasses current state-of-the-art methods with fewer updates, less data, and significantly shorter adaptation times. Compared with a previous SOTA SAR, we achieve 7.4% accuracy improvement and 4.2x faster adaptation speed on ImageNet-C.

Large language models (LLMs) have achieved remarkable success in machine translation, demonstrating impressive performance across diverse languages. However, translationese—characterized by overly literal and unnatural translations—remains a persistent challenge in LLM-based translation systems. Despite their pre-training on vast corpora of natural utterances, LLMs exhibit translationese errors and generate unexpected unnatural translations, stemming from biases introduced during supervised fine-tuning (SFT). In this work, we systematically evaluate the prevalence of translationese in LLM-generated translations and investigate its roots during supervised training. We introduce methods to mitigate these biases, including polishing golden references and filtering unnatural training instances. Empirical evaluations demonstrate that these approaches significantly reduce translationese while improving translation naturalness, validated by human evaluations and automatic metrics. Our findings highlight the need for training-aware adjustments to optimize LLM translation outputs, paving the way for more fluent and target-language-consistent translations.

In this paper, an output feedback adaptive fuzzy leader-following event-triggered consensus (ETC) problem for switched nonlinear multi-agent systems (SNMASs) is investigated. In the considered systems, only outputs of leader and followers are measurable at sampling instants. To deal with this problem, a novel auxiliary system and observers are employed. Then, corresponding error systems are established and the consensus problem is transformed into a stabilization problem. To achieve better approximation ability for nonlinear uncertainties, the interval type-2 fuzzy logic systems (IT2FLSs) are utilized. In order to reduce the wastage of communication resources (WCRs), an adaptive fuzzy ETC protocol with a discrete-time switching threshold event-triggering mechanism (ETM) which is monitored only at sampling instants is provided. It is proved that the output feedback adaptive fuzzy ETC protocol enables the realization of the consensus target. Finally, to verify the effectiveness of the proposed scheme, the ETC protocol is applied to a practical example.

Test-time adaptation (TTA) aims to boost the generalization capability of a trained model by conducting self-/unsupervised learning during the testing phase. While most existing TTA methods for video primarily utilize visual supervisory signals, they often overlook the potential contribution of inherent audio data. To address this gap, we propose a novel approach that incorporates audio information into video TTA. Our method capitalizes on the rich semantic content of audio to generate audio-assisted pseudo-labels, a new concept in the context of video TTA. Specifically, we propose an audio-to-video label mapping method by first employing pre-trained audio models to classify audio signals extracted from videos and then mapping the audio-based predictions to video label spaces through large language models, thereby establishing a connection between the audio categories and video labels. To effectively leverage the generated pseudo-labels, we present a flexible adaptation cycle that determines the optimal number of adaptation iterations for each sample, based on changes in loss and consistency across different views. This enables a customized adaptation process for each sample. Experimental results on two widely used datasets (UCF101-C and Kinetics-Sounds-C), as well as on two newly constructed audio-video TTA datasets (AVE-C and AVMIT-C) with various corruption types, demonstrate the superiority of our approach. Our method consistently improves adaptation performance across different video classification models and represents a significant step forward in integrating audio information into video TTA. The code and datasets will be made publicly available.

This article presents a discussion on the adaptive fuzzy fully distributed consensus problem of heterogeneous switched nonlinear multiagent systems (SNMASs). The considered agents contain both first- and second-order dynamics. Fuzzy logic systems are introduced to approximate the unknown nonlinear terms of the SNMASs. Influenced by the systems' own factors or the external environment, the subsystems of agents may be unstabilizable and existing control strategies may not be able to ensure the stability of the systems. Therefore, it is necessary to provide a new consensus protocol to address this problem. In this article, a fully distributed consensus protocol is provided that includes an auxiliary system and a series of state-dependent switching laws. To ensure the stability of whole SNMASs and realize the consensus objective, convex combination technology and the single Lyapunov function method are adopted. Finally, the effectiveness of the proposed scheme is verified through simulation results.

Industrial cyber–physical–social systems (CPSSs), a prominent data-driven paradigm, tightly couple and coordinate social space into cyber–physical systems (CPSs) within industrial environments. With the proliferation of cloud–fog computing, cloud–fog computing becomes the most prominent computing paradigm used to implement industrial data analysis. However, the open environment of cloud–fog computing and the limited control of industrial CPSSs users make industrial data analysis without compromising users’ privacy one great research challenge in practical cloud–fog-based industrial applications. High-order Bi-Lanczos (HOBI-Lanczos) approach has shown remarkable success in heterogeneous data analysis in industrial applications. In this article, a novel privacy preserving HOBI-Lanczos approach using tensor train in cloud–fog computing is proposed for industrial data applications. Specifically, a privacy preserving industrial data analysis model using cloud–fog computing and tensor train is firstly proposed. The proposed model enables fogs and clouds to securely carry out industrial data analysis for large-scale tensors given in a tensor train format. In addition, by using this model, a privacy preserving HOBI-Lanczos approach is provided. Last but not least, by using a brain-controlled robot system case study, the proposed approach is theoretically and empirically analyzed. Our proposed approach is proven to be secure. A series of experiments corroborate the superiority of the proposed approach in cloud–fog computing for industrial applications.

Tucker decomposition has been widely used to extract meaningful and underlying data from heterogeneous data generated by different kinds of devices in a wide range of industrial Internet of Things (IIoT) applications. IIoT data uploaded to the cloud contain personal and sensitive information; thus, there is a growing concern about data privacy. Current existing data analysis solutions, however, assume that the data are reliably and securely collected from different IIoT data providers, an assumption that is not always true in the real world. To address the issues, in this article we propose a privacy-preserving tucker train decomposition based on gradient descent over blockchain-based encrypted IIoT data. Specifically, we use blockchain techniques to enable IIoT data providers to reliably and securely share their data by encrypting them locally before recording them in the blockchain. We use tensor train (TT) theory to build an efficient TT-based tucker decomposition based on gradient descent that tremendously reduces the number of elements to be updated during the tucker decomposition. We utilize the massive resources of fogs and clouds to implement an efficient privacy-preserving tucker train decomposition scheme. We use homomorphic encryption to build our scheme that does complete tucker train decomposition without the involvement of users. Results from a series of extensive experiments on synthetic datasets and real-world datasets demonstrate that our proposed scheme is efficient.

Secure cyber-physical-social big data computations are being increasingly used to protect the users' data security in cyber-physical-social systems (CPSS). Despite the increasing popularity, how to process the tasks of the secure cyber-physical-social big data computations, while taking care of the energy consumption and meeting the users' requirements, remains challenging. To address the problem, in this work, we propose a novel tensor-based optimization model for the secure sustainable cyber-physical-social big data computations. The proposed model is a general and fine-grained model, which can jointly optimize the execution time, energy consumption, reliability, and quality of experience, and can comprehensively take into account step, task, time slot, type, node, core, cryptosystem, and security level. To our knowledge, this is the first study to holistically optimize the tasks in the secure cyber-physical-social big data computations. To illustrate the proposed model, the case study of the secure high-order Lanczos in cloud-assisted CPSS is presented. Finally, the proposed model is empirically evaluated by using multi-objective optimization, and the extensive results demonstrate that from the users' perspective our proposed tensor-based optimization model is preferable for the secure sustainable cyber-physical-social big data computations.

Smart environments, also referred to as cyber-physical-social systems (CPSSs), are expected to significantly benefit from the integration of edge, fog, and cloud for intelligence service flexibility, efficiency, and cost saving. High-order Bi-Lanczos method has emerged as a powerful tool serving as multi-dimensional data processing, such as prevailing feature extraction, classification, and clustering of high-order data, in CPSSs. However, integrated edge-fog-cloud architecture is open and users have very limited control; how to carry out big data processing without compromising the security and privacy is a challenging issue in edge-fog-cloud-assisted smart applications. In this work, we propose a novel and practical privacy-preserving high-order Bi-Lanczos scheme in integrated edge-fog-cloud architectural paradigm for smart environments. More precisely, we first propose a privacy-preserving big data processing model using the synergy of edge, fog, and cloud. The proposed model enables edge, fog, and cloud to cooperatively complete big data processing without compromising users’ privacy for large-scale tensor data in CPSSs. Subsequently, making use of the model, we present a privacy-preserving high-order Bi-Lanczos scheme. Finally, we theoretically and empirically analyze the security and efficiency of the proposed privacy-preserving high-order Bi-Lanczos scheme based on an intelligent surveillance system case study. And the results demonstrate that the proposed scheme provides a privacy-preserving and efficient way of computations in integrated edge-fog-cloud paradigm for smart environments.

Multiple clustering analysis has the clear advantages to discover latent data pattern in big data from different views, so it has tremendous practical values in automation industries. However, most of current algorithms are difficult to group heterogeneous data to multiple clusterings according to the requirements of different applications. This paper presents a flexible multiple clustering analytic and service framework, and a novel tensor-based multiple clusterings (TMC) approach. Heterogeneous data objects in cyber-physical-social systems are first represented as low-order tensors and a weight tensor construction approach is proposed to measure the importance of attributes combinations in heterogeneous feature spaces. Then, a selective weighted tensor distance is explored to cluster tensorized data objects for different applications. This paper, through a real-world smart bike maintenance system, illustrates TMC and evaluates its clustering performance. Experimental results reveal TMC can obtain higher quality clustering results but with lower redundancies to meet different requirements of applications in automation systems.