Xiamen University Malaysia

PurposeThis study aims to investigate the role of fashion influencers in shaping consumer behaviour within an omnichannel retail environment using the Malaysian market.Design/methodology/approachSurvey data were collected with a sample size of 298. To study trustworthiness, perceived expertise, similarity and attractiveness as predictors of consumer purchase intention, this study employed SmartPLS 4.0's structural equation modelling tool to analyse the data. Other than the usual reliability and validity tests, this study explores consumer attitudes as an important mediator in the structural model.FindingsThe results of the structural model indicate that the trustworthiness, perceived expertise, similarity and attractiveness of the fashion influencers had positive and significant effects on consumer purchase intention, with perceived expertise showing the strongest effect. The study also showed that consumer attitude plays a complete mediation role in the effect of trustworthiness on the purchase intention, as well as attractiveness on the purchase intention. Nonetheless, the effect of perceived expertise and similarity on purchase intention passed partially through consumer attitude.Research limitations/implicationsThis study found that consumer attitude plays a crucial role in mediating influencer attractiveness and trustworthiness on purchase intention. We highlight the importance of this dynamic in fashion marketing for the firms that are engaging fashion influencers.Originality/valueIn this study, we propose that fashion influencers serve as a key driver in shaping consumer preferences and purchasing behaviour. By employing empirical data and the partial least squares structural equation model, the study offers actionable insights for marketing strategists and fashion brand managers. The findings highlight effective strategies for leveraging fashion influencer marketing within the omnichannel retail framework to enhance consumer engagement and drive sales.

The potential for artificial intelligence (AI) to make a transformative impact in the chemistry research and education field begins with the contributions of university administrators, instructors, and students. As a team of international authors from 9 global regions representing these 3 roles in the university, we use public data to argue for a re-thinking and update to the chemistry curriculum to train the next generation of scientists. To prepare researchers for the future with AI, chemistry education must be designed for new approaches to problem-solving and ways of thinking. Analysis of higher education AI chemistry courses was extracted from publicly available information based on the 2024 QS World University Ranking and separated into two categories: “computational coursework” and “AI/ML coursework”. The data revealed that institutions with existing computational coursework also tend to offer AI/machine learning coursework. The paucity of these courses may reflect different funding priorities or a shortage of faculty trained in AI chemistry. Overall, immediate users such as university teachers can apply the Technological Pedagogical Content Knowledge (TPACK)-AI framework, originally proposed by Mishra and Koehler (2006) to describe the synergistic intersection of teachers’ technological, pedagogical, and content knowledge, and recently extended to encompass AI (Mishra et al., 2023), for instructional design, inspiring curriculum material and pedagogies to be conducive to learning chemistry. While novel approaches to teaching and learning can face significant implementation challenges due to various cultures, pedagogical innovation must continue to remain a priority to ensure our training remains relevant for future students. In this paper, we encourage the development of normative frameworks within higher education that avoid unequal workloads on academic staff and ensure all stakeholders have a voice in the creation of curriculum and training of our students for the future.

Let $(M,g)$ be a smooth Riemannian manifold, $K$ a compact Lie group and $p:P\to M$ a principal $K$-bundle over $M$ endowed with a connection $A$. Fixing a bi invariant inner product on Lie algebra $\mathfrak{k}$ of $K$, the connection $A$ and metric $g$ define a Riemannian metric $g_A$ on $P$. Let $\tilde {X}$ be the horizontal lift of vector field $X$ on $M$ and, let $\xi^\nu$ be the vertical field associated with section $\nu\in A^0(\mathrm{ad}( P))$ of the adjoint bundle. It is proved that the connection $A$ is invariant under the 1-parameter group of local diffeomorphism generated by $\tilde{ X}+\xi^\nu$ if and only if $X$ and $\nu$ satisfy the generalized moment map equation $\iota_XF_A=-\nabla^A\nu$. The Lie algebra of fiber preserving Killing fields of $(P,g_A)$ is studied, in the case where $K$ is compact, connected and semisimple.

We study the infinitesimal variation of Hodge structure for families of algebraic curves and extend the classical theory from smooth curves to singular and non–planar settings. Using the deformation space Ext^1(Ω_X,𝒪_X) and the dualizing sheaf, we define a singular analogue of maximal infinitesimal variation. For equisingular families of plane curves with planar Gorenstein singularities, we prove that the infinitesimal variation attains maximal rank equal to the arithmetic genus. We show that the rank decomposes into a geometric contribution from the normalization and a singular contribution measured by the δ–invariants. For non–equisingular degenerations, the rank defect equals the drop of the total δ–invariant and admits an interpretation in terms of vanishing cycles and mixed Hodge structures. We further extend the results to non–planar curves under suitable Petri and deformation conditions.

Offline reinforcement learning (RL) aims to learn effective agents from previously collected datasets, facilitating the safety and efficiency of RL by avoiding real-time interaction. However, in practical applications, the approximation error of the out-of-distribution (OOD) state-actions can cause considerable overestimation due to error exacerbation during training, finally degrading the performance. In contrast to prior works that merely addressed the OOD state-actions, we discover that all data introduces estimation error whose magnitude is directly related to data sparsity. Consequently, the impact of data sparsity is inevitable and vital when inhibiting the error exacerbation. In this article, we propose an offline RL approach to inhibit error exacerbation with data sparsity (IEEDS), which includes a novel value estimation method to consider the impact of data sparsity on the training of agents. Specifically, the value estimation phase includes two innovations: 1) replace Q-net with V-net, a smaller and denser state space makes data more concentrated, contributing to more accurate value estimation and 2) introduce state sparsity to the training by design state-aware-sparsity Markov decision process (MDP), further lessening the impact of sparse states. We theoretically prove the convergence of IEEDS under state-aware-sparsity MDP. Extensive experiments on offline RL benchmarks reveal that IEEDS's superior performance.