Audio-driven talking head synthesis aims to generate lifelike facial animations synchronized with audio. Current approaches primarily focus on lip motion information in 2D visual space for lip-audio synchronization and expressive lip dynamic, often neglecting 3D geometric motion representations of the lips that can more accurately capture lip movements in real-world scenarios. This oversight can result in suboptimal lip dynamic authenticity. In this work, we introduce a novel 3D Temporal Representation Learning (3D-TRL) algorithm that models 3D lip temporal information as latent representations and utilizes these representations as additional supervision to enhance dynamic authenticity. To achieve this, we leverage the geometric mesh constructed from the 3D Morphable Model (3DMM) as our 3D information of the lip and explore two self-supervised strategies to learn temporal representation in 3D geometric space. First, we propose a Reconstruction-Oriented 3D-TRL algorithm that reconstructs the input to obtain motion tokens in hidden space, encapsulating content while capturing richer contextual representations of the sequence. Second, we develop a Contrastive-Based 3D-TRL algorithm that utilizes contrastive learning to extract hidden 3D motion representations. This algorithm employs data augmentation strategies appropriate specifically for the 3D temporal sequences of the lips. Extensive experiments demonstrate that our approach, as a versatile and adaptable supervisory, can be integrated into various state-of-the-art network frameworks, leading to substantial enhancements in lip dynamic authenticity.

Digital Twin, the technological advance that has emerged in the Industry 4.0 era, will become one of the six technological pillars forming Industry 5.0, according to the vision of the European Commission. This promotion relies on the fact that digital-twin technologies are getting more mature and widely adopted in both the public and private sectors. Besides the progressions, a new challenge has arisen: harmonizing multiple digital twins in a single digital twin system. Indeed, digital twins built on different technologies addressing different aspects of an asset are not interoperable by default, making the interaction between them uneasy. Moreover, as an asset must have a connection with each digital twin for data exchange, its multiple connections with multiple digital twins may cause the asset itself and the involved network infrastructures to be overloaded. Regarding the above challenge, this paper proposes an approach based on the Asset Administration Shell (AAS) standard to harmonize multiple digital twins as a single digital twin with multiple facets.

Recent progress of deep generative models in the vision and language domain has stimulated significant interest in more structured data generation such as molecules. However, beyond generating new random molecules, efficient exploration and a comprehensive understanding of the vast chemical space are of great importance to molecular science and applications in drug design and materials discovery. In this paper, we propose a new framework, ChemFlow, to traverse chemical space through navigating the latent space learned by molecule generative models through flows. We introduce a dynamical system perspective that formulates the problem as learning a vector field that transports the mass of the molecular distribution to the region with desired molecular properties or structure diversity. Under this framework, we unify previous approaches on molecule latent space traversal and optimization and propose alternative competing methods incorporating different physical priors. We validate the efficacy of ChemFlow on molecule manipulation and single- and multi-objective molecule optimization tasks under both supervised and unsupervised molecular discovery settings. Codes and demos are publicly available on GitHub at [https://github.com/garywei944/ChemFlow](https://github.com/garywei944/ChemFlow).

Designing novel functional proteins crucially depends on accurately modeling their fitness landscape. Given the limited availability of functional annotations from wet-lab experiments, previous methods have primarily relied on self-supervised models trained on vast, unlabeled protein sequence or structure datasets. While initial protein representation learning studies solely focused on either sequence or structural features, recent hybrid architectures have sought to merge these modalities to harness their respective strengths. However, these sequence-structure models have so far achieved only incremental improvements when compared to the leading sequence-only approaches, highlighting unresolved challenges effectively leveraging these modalities together. Moreover, the function of certain proteins is highly dependent on the granular aspects of their surface topology, which have been overlooked by prior models. To address these limitations, we introduce the Sequence-Structure-Surface Fitness (**S3F**) model — a novel multimodal representation learning framework that integrates protein features across several scales. Our approach combines sequence representations from a protein language model with Geometric Vector Perceptron networks encoding protein backbone and detailed surface topology. The proposed method achieves state-of-the-art fitness prediction on the ProteinGym benchmark encompassing 217 substitution deep mutational scanning assays, and provides insights into the determinants of protein function. Our code is at https://github.com/DeepGraphLearning/S3F.

As digital twins gain momentum in their usage in diverse domains, the concept of Asset Administration Shells (AAS) has become very relevant for achieving the digital twin approach, where Administration Shells are the digital representation of physical assets. Being a relatively new concept in the Industrial Internet of Things (IIoT) domain, the tools and approaches for creating and deploying AASs are likewise in infancy. This paper introduces an open-source tool, Papyrus4Manufacturing, which provides a model-based systems engineering approach to the AAS. This toolset supports the creation of AAS digital twins from modeling to automatic deployment and connection to assets using the OPC UA protocol. This paper also includes an evaluation of its usability, as it is put to test with an academic use case.

We introduce M$^2$Hub, a toolkit for advancing machine learning in materials discovery. Machine learning has achieved remarkable progress in modeling molecular structures, especially biomolecules for drug discovery. However, the development of machine learning approaches for modeling materials structures lag behind, which is partly due to the lack of an integrated platform that enables access to diverse tasks for materials discovery. To bridge this gap, M$^2$Hub will enable easy access to materials discovery tasks, datasets, machine learning methods, evaluations, and benchmark results that cover the entire workflow. Specifically, the first release of M$^2$Hub focuses on three key stages in materials discovery: virtual screening, inverse design, and molecular simulation, including 9 datasets that covers 6 types of materials with 56 tasks across 8 types of material properties. We further provide 2 synthetic datasets for the purpose of generative tasks on materials. In addition to random data splits, we also provide 3 additional data partitions to reflect the real-world materials discovery scenarios. State-of-the-art machine learning methods (including those are suitable for materials structures but never compared in the literature) are benchmarked on representative tasks. Our codes and library are publicly available at \url{https://github.com/yuanqidu/M2Hub}.

Lot-size-one systems as well as plug and produce concepts imply (1) producing increased variety of products in a highly flexible and timely manner, and (2) making commissioning and maintenance more flexible. The speed with which manufacturers, in particular SMEs, can reconfigure the production to a new run and thus respond to clients and avoid costly machine downtime is critical to maintaining commercial success and profit margins. The manufacturing systems of tomorrow must offer a high degree of autonomy, be quickly re-planned to other operations, and cope with a wide variety of unforeseen situations, in a secure and safe manner. In this context, the Asset Administration Shell (AAS) is an emergent standard that leverages the digital twin approach and provides concepts for describing capabilities and skills of I4.0 components in order to automate the reconfiguration process. This article proposes a capability-based operation and engineering approach to tackle the syntactic and semantic interoperability problems in flexible production lines. We demonstrate the implementation of the AAS standard in the open source model-driven workbench Papyrus; then we assess its usability for modeling a production cell use case in order to implement a capability-based reconfiguration approach for flexible production lines.

The future intelligent manufacturing systems should possess a high degree of autonomy, which is able to monitor the entire production process, quickly re-plan operations, and respond to various unforeseen situations in a secure and safe manner. This can achieve rapid response to customers and avoid costly machine downtime, which is crucial to maintaining business success and profitability. The Asset Administration Shell (AAS) is an emerging standard in the I4.0 (Industry 4.0) domain. Based on the concept of digital twins, it provides concepts for describing the digital representation of I4.0 assets including their capabilities and skills. The AAS provides also responses to the challenge of syntactic and semantic interoperability that the flexible and autonomous production lines are facing. In this article, we propose a capability-based operation and engineering approach for flexible production lines. Our approach is relying on the AAS standard which is a very wide and rich specification. Consequently, we describe the subset of AAS modelling concepts necessary for our approach, we clarify their semantics and we show their usage through a production cell use case. Furthermore, we demonstrate how these modelling concepts were tooled as an extension of the open source model-driven workbench Papyrus.

Adverse Drug Reactions (ADRs) has become a worldwide problem that draws the attention of people from all racial and ethnic groups. The number of deaths caused by ADRs has greatly increased and led to many drug withdrawals in the last decades. Recent research findings indicate that most ADRs can be effectively prevented to some extent by using computer-aided information technologies. Though many spontaneous reporting systems (SRSs) have been built to enhance the pharma-covigilance, the ADRs data is still very sparse because the large amount of reports obtained from consumers contains insufficient hints to identify a possible causal relationship between an adverse event and drug. Based on this motivation, we developed Adverse-Tracking, a spontaneous reporting system of ADRs via crowd-sourcing. Our proposed system interacts with consumers through a Q&A interface and collects the ADR reports. The decision tree support vector machine (DTSVM) based on the genetic algorithm is used in our system to automate the Q&A procedure. We carried out experiments to evaluate the performance at Peking University First Hospital. As demonstrated by the results, our system is an efficient tool to track and discover adverse events in the consumers' reports of ADRs, which facilitates the detection of “signal”.