Sequence Of Interactions Research Articles

Learning Robust Sequential Recommenders through Confident Soft Labels

Sequential recommenders that are trained on implicit feedback are usually learned as a multi-class classification task through softmax-based loss functions on one-hot class labels. However, one-hot training labels are sparse and may lead to biased training and sub-optimal performance. Dense, soft labels have been shown to help improve recommendation performance. However, how to generate high-quality and confident soft labels from noisy sequential interactions between users and items is still an open question. We propose a new learning framework for sequential recommenders, CSRec, which introduces c onfident s oft labels to provide robust guidance when learning from user-item interactions. CSRec contains a teacher module that generates high-quality and confident soft labels and a student module that acts as the target recommender and is trained on the combination of dense, soft labels and sparse, one-hot labels. We propose and compare three approaches to constructing the teacher module: (i) model-level, (ii) data-level, and (iii) training-level. To evaluate the effectiveness and generalization ability of CSRec, we conduct experiments using various state-of-the-art sequential recommendation models as the target student module on four benchmark datasets. Our experimental results demonstrate that CSRec is effective in training better-performing sequential recommenders.

ChatGPT: A canary in the coal mine or a parrot in the echo chamber? Detecting fraud with LLM: The case of FTX

Does the paradigm shift brought by Large Language Models (LLMs) hold the promise of revolutionizing financial analysis? Our article tackles this question by exploring fraud detection in cryptocurrency exchanges, with a focus on FTX. We study the abilities of generative artificial intelligence tools like ChatGPT to serve as early-warning systems of fraud and identify red flags in the particular and difficult case where no financial information is available. We recognize several challenges to provide insights beyond human knowledge. To achieve a higher degree of scrutiny, we highlight the role of sequential interactions between the AI Chatbot and the researcher as well as the inclusion of external contents, a technique known as Retrieval Augmented Generation (RAG). Therefore, this article serves as a cautionary tale on the necessary conditions to achieve augmented intelligence.

Advising Entrepreneurs: Optimal Recommendation of Alternatives

Problem definition: Facing emergent business challenges, entrepreneurs often seek guidance from experienced advisors. When multiple alternatives could potentially solve the entrepreneur’s problem, advisors can lead the entrepreneur’s exploration by choosing which alternative(s) to suggest and in what sequence. Methodology/results: We develop a dynamic game-theoretic model that captures the sequential interaction between an advisor and an entrepreneur. The advisor chooses how to recommend alternative solutions, and the entrepreneur chooses which solution to try. The trial’s success depends on the viability of a solution and the entrepreneur’s execution capability. When a trial of a recommended solution fails, the belief about the viability of the solution is updated. Managerial implications: Our analysis reveals that the advisor should strategically recommend alternatives based on the entrepreneur’s execution capability, trial costs, and correlation between alternatives (among other factors). When the trial of the first alternative fails, the advisor should readily offer a new alternative if the entrepreneur’s capability is either very high or very low. Otherwise, the advisor should encourage the entrepreneur to try the same solution multiple times. In order to motivate and sustain the entrepreneur’s exploration over time and across solutions, the advisor may find it optimal to recommend inferior solutions before superior ones (e.g., when trial costs are different or the entrepreneur can improve her capability with experience) or recommend multiple solutions simultaneously (e.g., when there is correlation between alternatives). Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2022.0361 .

An improved cross-domain sequential recommendation model based on intra-domain and inter-domain contrastive learning

Cross-domain recommendation aims to integrate data from multiple domains and introduce information from source domains, thereby achieving good recommendations on the target domain. Recently, contrastive learning has been introduced into the cross-domain recommendations and has obtained some better results. However, most cross-domain recommendation algorithms based on contrastive learning suffer from the bias problem. In addition, the correlation between the user’s single-domain and cross-domain preferences is not considered. To address these problems, a new recommendation model is proposed for cross-domain scenarios based on intra-domain and inter-domain contrastive learning, which aims to obtain unbiased user preferences in cross-domain scenarios and improve the recommendation performance of both domains. Firstly, a network enhancement module is proposed to capture users’ complete preference by applying a graphical convolution and attentional aggregator. This module can reduce the limitations of only considering user preferences in a single domain. Then, a cross-domain infomax objective with noise contrast is presented to ensure that users’ single-domain and cross-domain preferences are correlated closely in sequential interactions. Finally, a joint training strategy is designed to improve the recommendation performances of two domains, which can achieve unbiased cross-domain recommendation results. At last, extensive experiments are conducted on two real-world cross-domain scenarios. The experimental results show that the proposed model in this paper achieves the best recommendation results in comparison with existing models.

Multi-Scale Higher-Order Dependencies (MSHOD): Higher-Order Interactions Mining and Key Nodes Identification for Global Liner Shipping Network

Liner shipping accounts for over 80% of the global transportation volume, making substantial contributions to world trade and economic development. To advance global economic integration further, it is essential to link the flows of global liner shipping routes with the complex system of international trade, thereby supporting liner shipping as an effective framework for analyzing international trade and geopolitical trends. Traditional methods based on first-order global liner shipping networks, operating at a single scale, lack sufficient descriptive power for multi-variable sequential interactions and data representation accuracy among nodes. This paper proposes an effective methodology termed “Multi-Scale Higher-Order Dependencies (MSHOD)” that adeptly reveals the complexity of higher-order interactions among multi-scale nodes within the global liner shipping network. The key step of this method is to construct high-order dependency networks through multi-scale attributes. Based on the critical role of high-order interactions, a method for key node identification has been proposed. Experiments demonstrate that, compared to other methods, MSHOD can more effectively identify multi-scale nodes with regional dependencies. These nodes and their generated higher-order interactions could have transformative impacts on the network’s flow and stability. Therefore, by integrating multi-scale analysis methods to mine high-order interactions and identify key nodes with regional dependencies, this approach provides robust insights for assessing policy implementation effects, preventing unforeseen incidents, and revealing regional dual-circulation economic models, thereby contributing to strategies for global, stable development.

Adult Verbal Responses Facilitating Child Vocalizations

Objectives: This study aimed to investigate whether adults in the vicinity of children respond differently to various child vocalization types within three-event sequences of “child vocalization-adult verbal response-child vocalization” occurring in everyday life. Additionally, this study examined the types of adult verbal responses that facilitate child vocalizations within the series of three-event sequential interactions between children and adults. Methods: Data was collected using the LENA (Language ENvironment Analysis) from 30 children aged 12 to 18 months and their families. Three-event sequences of interactions between children and adults were selected from twenty 5-minute recorded segments with the highest conversational turn counts. Child vocalizations before and after adult verbal responses were broadly classified into canonical babblings and non-canonical babblings, and then coded into subcategories for each babbling type. Adult verbal responses were classified into different pragmatic types. Results: Adults responded differently depending on children’s non-canonical and canonical babblings. Among pragmatic types of adult verbal responses, imitation, recast, and expansion were prominently observed following canonical babblings with more than two syllables. Adult verbal responses such as acknowledgment and recast facilitated children in producing more advanced vocalization forms. Conclusion: This study supports bi-directional relationships in interactions between children and adults and provides insights into which child-adult vocal exchanges facilitate child vocalizations and early speech-language development.

Optimizing decision-making in uncertain environments through analysis of stochastic stationary Multi-Armed Bandit algorithms

Reinforcement learning traditionally plays a pivotal role in artificial intelligence and various practical applications, focusing on the interaction between an agent and its environment. Within this broad field, the multi-armed bandit (MAB) problem represents a specific subset, characterized by a sequential interaction between a learner and an environment where the agents actions do not alter the environment or reward distributions. MABs are prevalent in recommendation systems and advertising and are increasingly applied in sectors like agriculture and adaptive clinical trials. The stochastic stationary bandit problem, a fundamental category of MAB, is the primary focus of this article. Here, we delve into the implementation and analytical comparison of several key bandit algorithmsincluding Explore-then-Commit (ETC), Upper Confidence Bound (UCB), Thompson Sampling (TS), Epsilon-Greedy (-Greedy), SoftMax, and Conservative Lower Confidence Bound (CON-LCB)across various datasets. These datasets vary in the number of options (arms), reward distributions, and specific parameters, offering a broad testing ground. Additionally, this work provides an overview of the diverse applications of bandit problems across different fields, highlighting their versatility and broad impact.

Formation and Evolution of the Pacific‐North American (San Andreas) Plate Boundary: Constraints From the Crustal Architecture of Northern California

AbstractThe northward migration of the Mendocino triple junction (MTJ) drives a fundamental plate boundary transformation from convergence to translation; producing a series of strike‐slip faults, that become the San Andreas plate boundary. We find that the 3‐D structure of the Pacific plate lithosphere in the vicinity of the MTJ controls the location of San Andreas plate boundary formation. At the time of initiation of the Pacific‐North America plate boundary (∼30 Ma), the sequential interaction with the western margin of North America of the Pioneer Fracture Zone, soon followed by the Mendocino Fracture Zone, led to the capture of a small segment of partially subducted Farallon lithosphere by the Pacific plate, termed the Pioneer Fragment (PF). Since that time, the PF has translated with the Pacific Plate along the western margin of North America. Recently developed, high‐resolution seismic‐tomographic imagery of northern California indicates that (a) the PF is extant, occupying the western half of the slab window, immediately south of the MTJ; (b) the eastern edge of the PF lies beneath the newly forming Maacama fault system, which develops to become the locus for the primary plate boundary structure after approximately 6–10 Ma; and (c) the location of the translating PF adjacent to the asthenosphere of the slab window generates a shear zone within and below the crust that develops into the plate boundary faults. As a result, the San Andreas plate boundary forms interior to the western margin of North America, rather than at its western edge.

Unraveling the role of MgO in the Ru-Ba/MgO catalyst for boosting ammonia synthesis: Comparative study of MgO and MgAlOx supports

The growing emphasis on decarbonization and the use of renewable energy has sparked considerable interest in ammonia (NH3) synthesis under mild conditions, accompanied by the concurrent development of catalysts. Among these catalysts, BaO-promoted Ru has often been reported as a highly efficient catalyst for this reaction. Despite MgO being traditionally regarded as an optimal support for this catalytic system, its precise role has remained ambiguous. This study aims to elucidate the specific role of MgO as a support for Ru-BaO catalysts. Using pure MgO and MgO-Al2O3 mixed oxide (MgAlOx) with different MgO contents (MgO wt% = 5, 30, and 70) as supports, catalysts impregnated with equal amounts of Ba and Ru were prepared and analyzed to explore their NH3 synthesis activity and structural properties. Our observations reveal that while MgO does not directly promote Ru, it maintains robust basicity of BaO, enabling BaO to effectively function as an electronic promoter. In contrast, BaO supported on AlOx domain shows suppressed basicity, resulting in inertness as a promoter. Consequently, the Ru-Ba/MgO catalyst forms highly active Ru-BaO-MgO interfacial sites, leading to significantly higher activity, lower activation energy, and a distinct reaction mechanism compared to Ru-Ba/MgAlOx catalysts. This work sheds light on the importance of sequential interactions, i.e. support–promoter and promoter–active center interactions, in facilitating catalytic activity, thereby providing valuable insights essential for the design of advanced catalysts.

Calcium ions promote migrasome formation via Synaptotagmin-1.

Migrasomes, organelles crucial for cell communication, undergo distinct stages of nucleation, maturation, and expansion. The regulatory mechanisms of migrasome formation, particularly through biological cues, remain largely unexplored. This study reveals that calcium is essential for migrasome formation. Furthermore, we identify that Synaptotagmin-1 (Syt1), a well-known calcium sensor, is not only enriched in migrasomes but also indispensable for their formation. The calcium-binding ability of Syt1 is key to initiating migrasome formation. The recruitment of Syt1 to migrasome formation sites (MFS) triggers the swelling of MFS into unstable precursors, which are subsequently stabilized through the sequential recruitment of tetraspanins. Our findings reveal how calcium regulates migrasome formation and propose a sequential interaction model involving Syt1 and Tetraspanins in the formation and stabilization of migrasomes.

Criminals Can Be Leaders: An Endogenous Timing Approach to Security Games

Security Games traditionally assume sequential interactions, with the defender acting as the leader and the attacker as the follower. In this paper, we challenge such an assumption by showing that security interactions can also be simultaneous or that even the attacker might be the leader. Specifically, we consider a simple security game framework consisting of two possible targets, one attacker and one defender. Then, we extend that basic interaction using a well-known endogenous timing framework (the Game with Observable Delay) to endogenously obtain the timing of the game as part of the equilibrium outcome. Our results show that, for the same set of general assumptions, the interaction can be simultaneous or sequential (with either the defender or attacker being the leader). Although the simultaneous case could be discarded using payoff dominance, it is impossible to compare the two sequential configurations using the same criterion. The risk dominance criterion also turns out to be helpless in this case. Thus, we argue that considering attackers as leaders in Security Games is a plausible scenario that merits further consideration, not only from a theoretical point of view but also due to its alignment with certain types of real-world crime.

Current Insights into the Maturation of Epstein-Barr Virus Particles.

The three subfamilies of herpesviruses (alphaherpesviruses, betaherpesviruses, and gammaherpesviruses) appear to share a unique mechanism for the maturation and egress of virions, mediated by several budding and fusion processes of various organelle membranes during replication, which prevents cellular membrane disruption. Newly synthesized viral DNA is packaged into capsids within the nucleus, which are subsequently released into the cytoplasm via sequential fusion (primary envelopment) and budding through the inner and outer nuclear membranes. Maturation concludes with tegumentation and the secondary envelopment of nucleocapsids, which are mediated by budding into various cell organelles. Intracellular compartments containing mature virions are transported to the plasma membrane via host vesicular trafficking machinery, where they fuse with the plasma membrane to extracellularly release mature virions. The entire process of viral maturation is orchestrated by sequential interactions between viral proteins and intracellular membranes. Compared with other herpesvirus subfamilies, the mechanisms of gammaherpesvirus maturation and egress remain poorly understood. This review summarizes the major findings, including recently updated information of the molecular mechanism underlying the maturation and egress process of the Epstein-Barr virus, a ubiquitous human gammaherpesvirus subfamily member that infects most of the population worldwide and is associated with a number of human malignancies.

A hydrophobic groove in secretagogin allows for alternate interactions with SNAP-25 and syntaxin-4 in endocrine tissues

Vesicular release of neurotransmitters and hormones relies on the dynamic assembly of the exocytosis/trans-SNARE complex through sequential interactions of synaptobrevins, syntaxins, and SNAP-25. Despite SNARE-mediated release being fundamental for intercellular communication in all excitable tissues, the role of auxiliary proteins modulating the import of reserve vesicles to the active zone, and thus, scaling repetitive exocytosis remains less explored. Secretagogin is a Ca2+-sensor protein with SNAP-25 being its only known interacting partner. SNAP-25 anchors readily releasable vesicles within the active zone, thus being instrumental for 1st phase release. However, genetic deletion of secretagogin impedes 2nd phase release instead, calling for the existence of alternative protein-protein interactions. Here, we screened the secretagogin interactome in the brain and pancreas, and found syntaxin-4 grossly overrepresented. Ca2+-loaded secretagogin interacted with syntaxin-4 at nanomolar affinity and 1:1 stoichiometry. Crystal structures of the protein complexes revealed a hydrophobic groove in secretagogin for the binding of syntaxin-4. This groove was also used to bind SNAP-25. In mixtures of equimolar recombinant proteins, SNAP-25 was sequestered by secretagogin in competition with syntaxin-4. Kd differences suggested that secretagogin could shape unidirectional vesicle movement by sequential interactions, a hypothesis supported by in vitro biological data. This mechanism could facilitate the movement of transport vesicles toward release sites, particularly in the endocrine pancreas where secretagogin, SNAP-25, and syntaxin-4 coexist in both α- and β-cells. Thus, secretagogin could modulate the pace and fidelity of vesicular hormone release by differential protein interactions.

A reinforcement learning framework for improving parking decisions in last-mile delivery

This study leverages simulation-optimisation with a Reinforcement Learning (RL) model to analyse the routing behaviour of delivery vehicles (DVs). We conceptualise the system as a stochastic k-armed bandit problem, representing a sequential interaction between a learner (the DV) and its surrounding environment. Each DV is assigned a random number of customers and an initial delivery route. If a loading zone is unavailable, the RL model is used to select a delivery strategy, thereby modifying its route accordingly. The penalty is gauged by the additional trucking and walking time incurred compared to the originally planned route. Our methodology is tested on a simulated network featuring realistic traffic conditions and a fleet of DVs employing four distinct lastmile delivery strategies. The results of our numerical experiments underscore the advantages of providing DVs with an RL-based decision support system for en-route decision-making, yielding benefits to the overall efficiency of the transport network. Highlights Combining simulation and optimisation algorithms with reinforcement learning Model DVs en-route parking decisions with a k-armed bandit algorithm Evaluating the impacts of delivery strategies on traffic congestion and in last-mile delivery efficiency

Transition Propagation Graph Neural Networks for Temporal Networks.

Researchers of temporal networks (e.g., social networks and transaction networks) have been interested in mining dynamic patterns of nodes from their diverse interactions. Inspired by recently powerful graph mining methods like skip-gram models and graph neural networks (GNNs), existing approaches focus on generating temporal node embeddings sequentially with nodes' sequential interactions. However, the sequential modeling of previous approaches cannot handles the transition structure between nodes' neighbors with limited memorization capacity. In detail, an effective method for the transition structures is required to both model nodes' personalized patterns adaptively and capture node dynamics accordingly. In this article, we propose a method, namely t ransition p ropagation g raph n eural n etworks (TIP-GNN), to tackle the challenges of encoding nodes' transition structures. The proposed TIP-GNN focuses on the bilevel graph structure in temporal networks: besides the explicit interaction graph, a node's sequential interactions can also be constructed as a transition graph. Based on the bilevel graph, TIP-GNN further encodes transition structures by multistep transition propagation and distills information from neighborhoods by a bilevel graph convolution. Experimental results over various temporal networks reveal the efficiency of our TIP-GNN, with at most 7.2% improvements of accuracy on temporal link prediction. Extensive ablation studies further verify the effectiveness and limitations of the transition propagation module. Our code is available at https://github.com/doujiang-zheng/TIP-GNN.

A mobile node path optimization approach based on Q-learning to defend against cascading failures on static-mobile networks

The research on cascading failures in static networks has become relatively mature, and an increasing number of scholars have started to explore the network scenarios where mobile nodes and static nodes coexist. In order to enhance the resilience of static-mobile networks against cascading failures, an algorithm based on Q-learning for optimizing the path of the mobile node is proposed in this paper. This paper proposes a Q-learning-based algorithm for optimizing the path of the mobile node. To achieve this objective, a cascading failure model is established based on sequential interactions between mobile nodes and static nodes in this study. In this model, the motion paths of the mobile node are generated by the Q-learning algorithm. Based on this approach, extensive experiments are conducted, and the results demonstrate the following findings: 1) By increasing the adjustable load parameters of static nodes in the network, the occurrence of cascading failures is delayed, and the frequency of cascading failures is decreased. 2) Increasing the adjustable load parameter, capacity parameter, and network size of static nodes contributes to the network's resilience against cascading failures. 3) As the communication radius of the mobile node increases, the scale of failures in the static network initially increases and then decreases. 4) Different trajectories of the mobile node have a significant impact on network robustness, and paths generated based on Q-learning algorithm exhibit significantly improved network robustness compared to Gaussian-Markov mobility trajectories. The Q-learning algorithm is compared to the Ant Colony Optimization algorithm in terms of execution time, path length, and network robustness, and the Q-learning algorithm demonstrating favorable performance. These experimental results can be valuable for theoretical research on cascading failures in static-mobile networks.

Collaborative game-based task offloading scheme in the UAV-TB-assisted battlefield network platform

In the sixth-generation (6G) wireless networks, the use of unmanned aerial vehicles (UAVs) and tethered balloons (TBs) to assist cellular networks has attracted considerable attentions due to their dynamic and quick deployment with their relative low cost. In this article, we propose a new task offloading scheme for smart devices in the modern battlefield area. By the integrative platform of TBs, UAVs and smart devices, the main challenges are (i) providing a task splitting algorithm for the partial offloading service, and (ii) develop a TB resource sharing algorithm to handle the offloading requests. For convenient wireless communications, UAVs work as relay nodes between TBs and individual devices. To achieve a mutually desirable solution, our proposed scheme is formulated as cooperative game models. First, the sequential Raiffa bargaining solution is applied to split the computation-intensive task of each smart device in the battlefield area. Second, the average-surplus value is adopted to effectively share the TB computing resource. Based on the reciprocal combination of two cooperative game solutions, we explore the sequential interaction of TBs, UAVs and battlefield devices, and jointly design our integrated control scheme for offloading services. According to the synergy effect, our hybrid approach can provide a fair-efficient solution in the UAV-TB-assisted battlefield network infrastructure. Finally, extensive simulations are conducted, and the results demonstrate the superiority of our proposed scheme over the existing baseline protocols.

Insight into the Effect of Submicellar Concentrations of Sodium Deoxycholate on the Structure, Stability, and Activity of Bovine and Human Serum Albumin: An Interesting Comparison between Single and Double Tryptophan Proteins.

The progressive escalation in the applications of bile salts in diverse fields has triggered research on their interaction with various biological macromolecules, especially with proteins. A proper understanding of the interaction process of bile salts, particularly in the lower concentrations range, with the serum albumin seems important since the normal serum concentration of bile salts is approximately in the micromolar range. The current study deals with a comprehensive and comparative analysis of the interaction of submicellar concentrations of sodium deoxycholate (NaDC) with two homologous transport proteins: bovine serum albumin (BSA) and human serum albumin (HSA). HSA and BSA with one and two tryptophans, respectively, provide the opportunity for an interesting comparison of tryptophan fluorescence behavior on interaction with NaDC. The study suggests a sequential interaction of NaDC in three discrete stages with the two proteins. A detailed study using warfarin and ibuprofen as site markers provides information about the sites of interaction, which is further confirmed by inclusive molecular dynamics simulation analysis. Moreover, the comparison of the thermodynamics and stability of the NaDC-serum albumin complexes confirms the stronger interaction of NaDC with BSA as compared to that with HSA. The differential interaction between the bile salt and the two serum albumins is further established from the difference in the extent of decrease in the esterase-like activity assay of the proteins in the presence of NaDC. Therefore, the present study provides important insight into the effect of submicellar concentrations of NaDC on the structure, stability, and activity of the two homologous serum albumins and thus can contribute not only to the general understanding of the complex nature of serum albumin-bile salt interactions but also to the design of more effective pharmaceutical formulations in the field of drug delivery and biomedical research.

Identification of a new export signal that targets early subunits to the flagellar type III secretion export machinery.

Many bacterial pathogens utilize T3SS to inject virulence proteins (effectors) into host cells or to assemble flagella on the bacterial cell surface. Bacterial flagella present a paradigm for how cells build and operate complex cell-surface "nanomachines." Efficient subunit targeting from the bacterial cytosol to type III secretion systems is essential for rapid assembly and secretion by T3SSs. Subunits are thought to dock at the export machinery before being unfolded and translocated into the export channel. However, little is known about how subunits dock at the export machinery and the events that occur post docking. Here, we identified a new export signal within the C-termini of subunits that is essential for targeting of subunits to the type III export machinery. We show that this new export signal and previously identified export signals are recognized separately and sequentially, revealing a pathway for subunit transit through the type III export machinery in which sequential recognition events carry out different roles at major steps in the export pathway.

An Optimizing Business Process: A Comprehensive Analysis of Amazon Inc.’s Information Architecture

This analysis studies Amazon Inc.’s information system design in detail using context, use cases, test cases, class, and sequence diagrams. Our research aims to improve business processes by better understanding the system’s interactions, functionality, and data structure. The representation of data structure and entity relationships in class diagrams provides insights into the system’s component organization. Sequence diagrams also provide a clear picture of the sequential interactions between system elements, showing potential bottlenecks and optimization opportunities. It provides unique insights into Amazon’s approaches to keeping a competitive edge in the e-commerce industry, and it may be used as a resource for firms looking to improve their own information systems and procedures.