Search

Motivation: Retrosynthesis planning poses a formidable challenge in the organic chemical industry, particularly in pharmaceuticals. Single-step retrosynthesis prediction, a crucial step in the planning process, has witnessed a surge in interest in recent years due to advancements in AI for science. Various deep learning-based methods have been proposed for this task in recent years, incorporating diverse levels of additional chemical knowledge dependency. Results: This paper introduces UAlign, a template-free graph-to-sequence pipeline for retrosynthesis prediction. By combining graph neural networks and Transformers, our method can more effectively leverage the inherent graph structure of molecules. Based on the fact that the majority of molecule structures remain unchanged during a chemical reaction, we propose a simple yet effective SMILES alignment technique to facilitate the reuse of unchanged structures for reactant generation. Extensive experiments show that our method substantially outperforms state-of-the-art template-free and semi-template-based approaches. Importantly, our template-free method achieves effectiveness comparable to, or even surpasses, established powerful template-based methods. Scientific contribution: We present a novel graph-to-sequence template-free retrosynthesis prediction pipeline that overcomes the limitations of Transformer-based methods in molecular representation learning and insufficient utilization of chemical information. We propose an unsupervised learning mechanism for establishing product-atom correspondence with reactant SMILES tokens, achieving even better results than supervised SMILES alignment methods. Extensive experiments demonstrate that UAlign significantly outperforms state-of-the-art template-free methods and rivals or surpasses template-based approaches, with up to 5% (top-5) and 5.4% (top-10) increased accuracy over the strongest baseline. [ABSTRACT FROM AUTHOR]

Modeling human mobility helps to understand how people are accessing resources and physically contacting with each other in cities, and thus contributes to various applications such as urban planning, epidemic control, and location-based advertisement. Next location prediction is one decisive task in individual human mobility modeling and is usually viewed as sequence modeling, solved with Markov or RNN-based methods. However, the existing models paid little attention to the logic of individual travel decisions and the reproducibility of the collective behavior of population. To this end, we propose a Causal and Spatial-constrained Long and Short-term Learner (CSLSL) for next location prediction. CSLSL utilizes a causal structure based on multi-task learning to explicitly model the "when→what→where", a.k.a. "time→activity→location" decision logic. We next propose a spatial-constrained loss function as an auxiliary task, to ensure the consistency between the predicted and actual spatial distribution of travelers' destinations. Moreover, CSLSL adopts modules named Long and Short-term Capturer (LSC) to learn the transition regularities across different time spans. Extensive experiments on three real-world datasets show promising performance improvements of CSLSL over baselines and confirm the effectiveness of introducing the causality and consistency constraints. The implementation is available at https://github.com/urbanmobility/CSLSL. [ABSTRACT FROM AUTHOR]

People are keen to share their geospatial locations to access social activities or services via mobile internet, which provides a new perspective for us to understand human mobility. However, the collection of individual mobility data arouses great concern about privacy among the public. Specifically, anonymized mobility records can be used to imply the routine behavior of individuals such that people can be re-identified even if they share footprints with different accounts or platforms. In this context, this study explored the probability of re-identifying anonymized users with advanced deep learning techniques, only leveraging their trajectories collected during a long time period. Such a user re-identification task realizes the identification of anonymous users by mining the characteristics of anonymous trajectories. Prevailing methods adopt deep sequential learning models, such as recurrent neural networks (RNNs), to capture the inherent similarity between any two trajectories, replacing classic statistical models. Despite that, RNN-like models usually fail in learning effective knowledge from longer sequences, such as one's visited locations in one week. To this end, we propose a novel model based on the Siamese Transformer network. The entire model comprises a discriminant module and retrieval module. The discriminant module uses the Transformer model to detect the characteristics of the trajectory and employs an improved attention mechanism to achieve similarity measurement between trajectories. The retrieval module helps the model deal with the matching between the anonymous trajectory and user by constructing a mapping relationship between users and locations. Extensive experiments on four real-world location-based social network datasets demonstrated that our method outperforms existing methods. [ABSTRACT FROM AUTHOR]

Federated medical relation extraction enables multiple clients to train a deep network collaboratively without sharing their raw medical data. To handle the heterogeneous label distribution across clients, most of the existing works enforce regularization between local and global models during updating. In this paper, we propose the concept of major classifier vectors, which are a group of classifier vectors that characterize the representation space of relation classes well. They are obtained by comparing the inter-classifier similarity between clients, which is an ensembling method that avoids the bias introduced by weighted aggregation. We propose an algorithm named FedCMC, which restricts the updating of local models by contrasting with major classifier vectors to avoid overfitting to the local label distribution by comparison with major classifier vectors. Extensive experiments show that FedCMC outperforms the other state-of-the-art federated learning (FL) algorithms on three medical relation extraction datasets. [ABSTRACT FROM AUTHOR]

Multi-Label Text Classifiction (MLTC) aims to assign the most relevant labels to each given text. Existing methods demonstrate that label dependency can help to improve the model's performance. However, the introduction of label dependency may cause the model to suffer from unwanted prediction bias. In this study, we attribute the bias to the model's misuse of label dependency, i.e., the model tends to utilize the correlation shortcut in label dependency rather than fusing text information and label dependency for prediction. Motivated by causal inference, we propose a CounterFactual Text Classifier (CFTC) to eliminate the correlation bias, and make causality-based predictions. Specifically, our CFTC first adopts the predict-then-modify backbone to extract precise label information embedded in label dependency, then blocks the correlation shortcut through the counterfactual de-bias technique with the help of the human causal graph. Experimental results on three datasets demonstrate that our CFTC significantly outperforms the baselines and effectively eliminates the correlation bias in datasets. [ABSTRACT FROM AUTHOR]

This study investigated the influence of band microstructure induced by centerline segregation on carbide precipitation behavior and toughness in an 80 mm-thick 1 GPa low-carbon low-alloy steel plate. The quarter-thickness (1/4t) and half-thickness (1/2t) regions of the plate exhibited similar ductility and toughness after quenching. After tempering, the 1/4t region exhibited ∼50% and ∼25% enhancements in both the total elongation and low-temperature toughness at −40°C, respectively, without a decrease in yield strength, whereas the toughness of the 1/2t region decreased by ∼46°%. After quenching, both the 1/4t and 1/2t regions exhibited lower bainite and lath martensite concentrations, but only the 1/2t region exhibited microstructure bands. Moreover, the tempered 1/4t region featured uniformly dispersed short rod-like M23C6 carbides, and spherical MC precipitates with diameters of ∼20–100 nm and <20 nm, respectively. The uniformly dispersed nanosized M23C6 carbides and MC precipitates contributed to the balance of high strength and high toughness. The band microstructure of the tempered 1/2t region featured a high density of large needle-like M3C carbides. The length and width of the large M3C carbides were ∼200–500 nm and ∼20–50 nm, respectively. Fractography analysis revealed that the high density of large carbides led to delamination cleavage fracture, which significantly deteriorated toughness. [ABSTRACT FROM AUTHOR]

In this study, Cr-/Mo-free Cu-bearing steel with the addition of 1.5 wt% Cu was designed to improve the strength and toughness of the coarse-grained heat-affected zone (CGHAZ) at high heat input. A comparative test with traditional EH690 steel was conducted to analyze the strengthening and toughening mechanism. As heat input increased from 50 to 200 kJ/cm, microstructure of the CGHAZ for Cu-bearing steel transformed from lath/granular bainite to quasi-polygonal ferrite, while EH690 steel contained granular bainite under these conditions. The phase field simulation results showed that the increase in heat input promoted the diffusion of Cu and precipitation of nanoscale Cu-rich particles. At heat inputs of 100 kJ/cm and 200 kJ/cm, Cu-rich particles with sizes of 3.8 nm and 7.5 nm precipitated in the CGHAZ, providing precipitation strengthening increment of 47 MPa and 167 MPa, respectively. The precipitation strengthening of nanoscale Cu-rich particles compensated for the strength loss caused by the ferritic transformation of Cu-bearing steel. Low carbon and Cr-/Mo-free for Cu-bearing steel reduced the content of M-A constituents, leading to the increase in crack propagation energy and toughness. Under heat input of 200 kJ/cm, tensile strength, yield strength and impact energy at -40 °C of CGHAZ for Cu-bearing steel were 975 MPa, 837 MPa and 63.7 J, respectively, exhibiting the good mechanical properties under high heat input condition. [ABSTRACT FROM AUTHOR]

AA7020 sheets with different Zr contents were successfully fabricated by twin-roll casting process. In order to investigate the effect of Zr content on AA7020 twin-roll cast sheets, the microstructure and mechanical properties were detected by optical microscope, scanning electron microscope, transmission electron microscope, hardness tester and universal testing machine. The results show that proper addition of Zr content can effectively improve surface cracks, refine grains, and mechanical properties of twin-roll casting sheet. Zr mainly exists in the form of Al3Zr in aluminum alloy. With increasing of Zr content, Al3Zr phase tends to aggregate, even become potential cause of crack formation at 0.83% of Zr content. When the content of Zr is 0.33%, the comprehensive properties of AA7020 twin-roll cast sheets reach an idea value, i.e., yield strength 234.9 MPa, tensile strength 254.3 MPa, elongation is 10.8%, and hardness 176 HV. [ABSTRACT FROM AUTHOR]

AA7020 twin-roll cast sheets were warm rolled with different processing parameters to investigate the evolution of cracks, microstructure and properties by a single variable method. The as-cast rolled and as-warm rolled sheets were examined by optical microscopy (OM), scanning electron microscopy (SEM), energy-dispersive spectroscopy (EDS), a Vickers hardness tester and an electronic universal testing machine. As the reduction ratio increased, the grains became finer, the mechanical properties improved, and the cracks gradually healed. The grain size, mechanical properties, and cracks first increased and then decreased with increasing warm rolling temperature. The cracks healed by occlusion between grains. When the warm rolling temperature was too high, the grains grew upward, the mechanical properties decreased, and the healed cracks left scars. The optimum process parameters were a warm rolling temperature of 350 °C and a reduction ratio of 80%. [ABSTRACT FROM AUTHOR]

The rapid spread of residential broadband connections and Internet-capable consumer devices in home networks has changed the landscape of Internet traffic. To gain a deep understanding of Internet traffic for home networks, this paper develops a traffic monitoring platform that collects and analyzes home network traffic via programmable home routers and traffic profiling servers. Using traffic data captured from real home networks, we present traffic characteristics in home networks and then apply principal component analysis to uncover temporal correlations among application ports. In light of prevalent unwanted traffic on the Internet, we characterize the intensity, sources, and port diversity of unwanted traffic toward home networks. To the best of our knowledge, this paper is the first study to characterize network traffic of Internet-capable devices from inside home network. [ABSTRACT FROM AUTHOR]

In this paper, the sharing schemes of multicast in survivable Wavelength-Division Multiplexed (WDM) networks are studied and the concept of Shared Risk Link Group (SRLG) is considered. While the network resources are shared by the backup paths, the sharing way is possible to make the backup paths selfish. This selfishness leads the redundant hops of the backup route and a large number of primary lightpaths to share one backup link. The sharing schemes, especially, the self-sharing and cross-sharing, are investigated to avoid the selfishness when computing the backup light-tree. In order to decrease the selfishness of the backup paths, it is important to make the sharing links fair to be used. There is a trade-off between the self-sharing and cross-sharing, which is adjusted through simulation to adapt the sharing degree of each sharing scheme and save the network resources. [ABSTRACT FROM AUTHOR]

An introduction is presented in which the editor discusses various reports within the issue on topics including different approaches in calculating the amount of access network cables and fibres and the number of required network elements for Fibre To The Home (FTTH) networks, protection in passive optical networks (PONs) at minor investment cost, and problems related to the optical systems for telecommunication applications.