Visible-Infrared Person Re-Identification (VI-ReID) that matches pedestrian images across visible and infrared modalities suffers from substantial modality discrepancies and intra-class variations. While existing methods typically address the modality gap via style alignment, they often lose identity-relevant semantics and overlook fine-grained inter-class nuances, such as body part contours and structural cues around the head, shoulders, or feet. To tackle these challenges, we propose an Identity-Compensated Style Distillation (ICSD) network that enforces cross-modality style consistency and enhances the discriminative power of modality-invariant features. Specifically, ICSD comprises two core components: (1) a Style Knowledge Distillation (SKD) module, which integrates Style Discrepancy Reduction (SDR) and Identity Knowledge Compensation (IKC) to align modality styles while preserving identity-relevant semantics; (2) an Identity Discrimination Amplification (IDA) module, which captures and enhances subtle inter-class differences by refining identity-specific cues, thereby facilitating more accurate discrimination between different pedestrians. Extensive experiments on three public benchmarks-SYSU-MM01, RegDB, and LLCM-demonstrate that ICSD consistently outperforms state-of-the-art methods, validating the effectiveness and complementarity of its components.

Unsupervised Visible-Infrared Person Re-Identification (USVI-ReID) aims to match images of the same individual across visible and infrared modalities without relying on identity annotations. This task is critical in Visual Internet of Things (VIoT) systems. Existing methods typically generate pseudo-labels through clustering and cross-modality associations. However, the quality of these pseudo-labels is often compromised by unstable clustering performance, modality discrepancies, and unreliable matching strategies, resulting in suboptimal accuracy. Therefore, obtaining more reliable and robust pseudo-labels remains challenging in this domain. To address these challenges, we propose a Text-Guided Multi-Round Learning (TGMRL) framework that enhances the reliability and robustness of cross-modality pseudo-label associations. TGMRL comprises two core components: (1) the Text Dual-Cross Similarity Matching (TDSM) module, which facilitates cross-modality cluster alignment by constructing both visual and text-based cluster centers and integrating their similarity matrices for more accurate correspondence; and (2) the Multi-Round Pseudo-Label Guidance (MRPG) module, which enhances label consistency by imposing temporal regularization across clustering iterations through both similarity and Intersection-Over-Union (IoU) based measures. Extensive experiments on two benchmark datasets demonstrate that TGMRL significantly improves the reliability of pseudo-labels and achieves state-of-the-art performance on the USVI-ReID task.

In this work, we consider monotone Riemannian Variational Inequality Problems (RVIPs), which encompass both Riemannian convex optimization and minimax optimization as particular cases. In Euclidean space, the last-iterates of both the extragradient (EG) and past extragradient (PEG) methods converge to the solution of monotone variational inequality problems at a rate of $O\left(\frac{1}{\sqrt{T}}\right)$ (Cai et al., 2022). However, analogous behavior on Riemannian manifolds remains open. To bridge this gap, we introduce the Riemannian extragradient (REG) and Riemannian past extragradient (RPEG) methods. We demonstrate that both exhibit $O\left(\frac{1}{\sqrt{T}}\right)$ last-iterate convergence and $O\left(\frac{1}{{T}}\right)$ average-iterate convergence, aligning with observations in the Euclidean case. These results are enabled by judiciously addressing the holonomy effect so that additional complications in Riemannian cases can be reduced and the Euclidean proof inspired by the performance estimation problem (PEP) technique or the sum-of-squares (SOS) technique can be applied again.

This work in progress belongs to the innovative practice category. Nowadays, generative AI (also known as GenAI) can produce novel data samples that closely resemble authentic datasets. The advent of large language models (LLMs), in particular, has caused a huge interest in utilizing GenAI within and beyond the realm of higher education. However, little about engineering students' views and behaviours related to knowing and knowledge when using GenAI, such as ChatGPT, is known. In this WIP, we have engaged a class of N = 37 engineering students taking a postgraduate course titled “Social Media Analytics”. They were required to write essays related to their course learning in the form of blog posts. They were required to use LLM tools, such as ChatGPT, to assist their writing processes. Their GenAI usage was guided by the cognitive-agent approach, Search Tree, Analyze and Repair, and Selection (STARS), while STARS was proposed by Kirk et al. in AAAI 2024 to extend and complement prompt engineering. In addition, the participants were invited to fill in the Epistemic Cognition Inventory (ECI) questionnaire to associate five aspects of epistemic cognition (EC) with their writing experience. It is confirmed in our results that students' EC, i.e., their beliefs related to knowledge and knowing, significantly predict their prompting engagement and academic performance. However, students' academic performance is found to be significantly and negatively associated with their preference for GenAI usage. Here, we have uncovered engineering students' EC when interacting with generative AI, an area where little has been known so far. Our findings also suggest that proper use of GenAI prompting might promote engineering students' EC and, therefore, engineering learning.

The rapid proliferation of artificial intelligence (AI) services gives rise to the development of federated learning (FL), enabling the cooperative learning among wireless devices (WDs) with only local model parameters communicated. Nevertheless, the current emergence of large AI models renders the existing FL approaches inefficient, due to the huge communication overhead. In this paper, we propose a novel over-the-air federated distillation (FD) framework by synergizing the strength of FL and knowledge distillation to avoid the heavy local model transmission. Instead of sharing model parameters, only WDs' model outputs, referred to as knowledge, are shared and aggregated over-the-air by exploiting the superposition property of the multiple-access channel. Accordingly, we study the communication-learning co-design in over-the-air FD, aiming to maximize the learning convergence rate while meeting the power constraints of the transceivers. The main challenge lies in the intractability of the learning performance analysis, as well as the non-convex nature and the optimization spanning the whole FD training period. To tackle this problem, we propose an efficient algorithm to jointly optimize the transmit power of the WDs, estimator for over-the-air aggregation, and receiver beamforming per training round. Numerical results demonstrate that the proposed over-the-air FD achieves significant communication overhead reduction, with only a slight compensation of testing accuracy compared to conventional FL benchmarks.

Recent years have witnessed the development of federated learning (FL) that allows wireless devices (WDs) to collaboratively learn a global model under the coordination of a parameter server without sharing their local datasets. To meet the communication efficiency and privacy requirements, over-the-air computation and differential privacy (DP) have been incorporated in FL by leveraging the signal-superposition property of multiple-access channels and using artificial noises to perturb local model updates, thereby preserving DP. In this paper, we propose an exploration into device sampling with replacement as a potential mechanism for augmenting the DP levels of WDs in over-the-air FL. In particular, we delve into the joint optimization of device sampling strategy, the number of training rounds, and over-the-air transceiver design. Our goal is to maximize the learning performance while ensuring each WD meets the DP requirement. The problem is challenging due to the intractable FL convergence rate and privacy losses under random sampling, coupled with the strong interconnection among mixed continuous and integer decision variables. To tackle this problem, we first analyze the learning convergence rate and privacy losses of WDs. The analysis allows us to derive the optimal transceiver design per round in closed forms. Then, we propose an efficient alternating optimization algorithm by deriving the optimal device sampling strategy and the number of training rounds in semi-closed forms. Our numerical results, based on real-world learning tasks, showcase the effectiveness of our proposed approach compared with representative baselines.

In this paper, we consider the sequential decision problem where the goal is to minimize the general dynamic regret on a complete Riemannian manifold. The task of offline optimization on such a domain, also known as a geodesic metric space, has recently received significant attention. The online setting has received significantly less attention, and it has remained an open question whether the body of results that hold in the Euclidean setting can be transplanted into the land of Riemannian manifolds where new challenges (e.g., curvature) come into play. In this paper, we show how to get optimistic regret bound on manifolds with non-positive curvature whenever improper learning is allowed and propose an array of adaptive no-regret algorithms. To the best of our knowledge, this is the first work that considers general dynamic regret and develops “optimistic” online learning algorithms which can be employed on geodesic metric spaces.

Recent years have witnessed the development of federated learning (FL) that allows wireless devices (WDs) to collaboratively learn a global model under the coordination of a parameter server without sharing local datasets. To meet the communication efficiency and privacy requirements, over-the-air computation and differential privacy (DP) are further incorporated in FL by leveraging the signal-superposition property of multiple-access channels, as well as artificial noises to perturb local model updates for DP preservation. In this paper, we consider the device sampling with replacement, as an amplifier for the DP levels of WDs, in differentially private over-the-air FL. Accordingly, we study the joint optimization of device sampling strategy and over-the-air transceiver design that maximizes the learning performance while satisfying the DP requirement of each WD. The problem is challenging due to the intractable FL convergence rate and privacy losses under the sampling randomness, and the strong coupling among mixed decision variables. To tackle this problem, we first derive the analytical learning convergence rate and privacy losses of WDs, based on which the optimal transceiver design and device sampling strategy are obtained in closed forms. Numerical results demonstrate the effectiveness of our proposed approach compared with representative baselines.

The projection operation is a critical component in a wide range of optimization algorithms, such as online gradient descent (OGD), for enforcing constraints and achieving optimal regret bounds. However, it suffers from computational complexity limitations in high-dimensional settings or when dealing with ill-conditioned constraint sets. Projection-free algorithms address this issue by replacing the projection oracle with more efficient optimization subroutines. But to date, these methods have been developed primarily in the Euclidean setting, and while there has been growing interest in optimization on Riemannian manifolds, there has been essentially no work in trying to utilize projection-free tools here. An apparent issue is that non-trivial affine functions are generally non-convex in such domains. In this paper, we present methods for obtaining sub-linear regret guarantees in online geodesically convex optimization on curved spaces for two scenarios: when we have access to (a) a separation oracle or (b) a linear optimization oracle. For geodesically convex losses, and when a separation oracle is available, our algorithms achieve $O(T^{\frac{1}{2}})$, $O(T^{\frac{3}{4}})$ and $O(T^{\frac{1}{2}})$ adaptive regret guarantees in the full information setting, the bandit setting with one-point feedback and the bandit setting with two-point feedback, respectively. When a linear optimization oracle is available, we obtain regret rates of $O(T^{\frac{3}{4}})$ for geodesically convex losses and $O(T^{\frac{2}{3}}\log T)$ for strongly geodesically convex losses.

First-order optimization methods tend to inherently favor certain solutions over others when minimizing a given training objective with multiple local optima. This phenomenon, known as \emph{implicit bias}, plays a critical role in understanding the generalization capabilities of optimization algorithms. Recent research has revealed that gradient-descent-based methods exhibit an implicit bias for the $\ell_2$-maximal margin classifier in the context of separable binary classification. In contrast, generic optimization methods, such as mirror descent and steepest descent, have been shown to converge to maximal margin classifiers defined by alternative geometries. However, while gradient-descent-based algorithms demonstrate fast implicit bias rates, the implicit bias rates of generic optimization methods have been relatively slow. To address this limitation, in this paper, we present a series of state-of-the-art implicit bias rates for mirror descent and steepest descent algorithms. Our primary technique involves transforming a generic optimization algorithm into an online learning dynamic that solves a regularized bilinear game, providing a unified framework for analyzing the implicit bias of various optimization methods. The accelerated rates are derived leveraging the regret bounds of online learning algorithms within this game framework.

The classical algorithms for online learning and decision-making have the benefit of achieving the optimal performance guarantees, but suffer from computational complexity limitations when implemented at scale. More recent sophisticated techniques, which we refer to as $\textit{oracle-efficient}$ methods, address this problem by dispatching to an $\textit{offline optimization oracle}$ that can search through an exponentially-large (or even infinite) space of decisions and select that which performed the best on any dataset. But despite the benefits of computational feasibility, most oracle-efficient algorithms exhibit one major limitation: while performing well in worst-case settings, they do not adapt well to friendly environments. In this paper we consider two such friendly scenarios, (a) "small-loss" problems and (b) IID data. We provide a new framework for designing follow-the-perturbed-leader algorithms that are oracle-efficient and adapt well to the small-loss environment, under a particular condition which we call $\textit{approximability}$ (which is spiritually related to sufficient conditions provided in (Dudík et al., 2020)). We identify a series of real-world settings, including online auctions and transductive online classification, for which approximability holds. We also extend the algorithm to an IID data setting and establish a "best-of-both-worlds" bound in the oracle-efficient setting.

Among learning based hashing methods, supervised hashing seeks compact binary representation of the training data to preserve semantic similarities. Recent years have witnessed various problem formulations and optimization methods for supervised hashing. Most of them optimize a form of loss function with a regulization term, which can be viewed as a maximum a posterior (MAP) estimation of the hashing codes. However, these approaches are prone to overfitting unless hyperparameters are tuned carefully. To address this problem, we present a novel fully Bayesian treatment for supervised hashing problem, named Bayesian Supervised Hashing (BSH), in which hyperparameters are automatically tuned during optimization. Additionally, by utilizing automatic relevance determination (ARD), we can figure out relative discriminating ability of different hashing bits and select most informative bits among them. Experimental results on three real-world image datasets with semantic information show that BSH can achieve superior performance over state-of-the-art methods with comparable training time.