arXiv subjects

Shimin Cai

Publications and source records attributed to Shimin Cai.

15 recordsLinked to original sources

arXiv2026

Stop Treating Collisions Equally: Qualification-Aware Semantic ID Learning for Recommendation at Industrial Scale

Semantic IDs (SIDs) are compact discrete representations derived from multimodal item features, serving as a unified abstraction for ID-based and generative recommendation. However, learning high-quality SIDs remains challenging due to two issues. (1) Collision problem: the quantized token space is prone to collisions, in which semantically distinct items are assigned identical or overly similar SID compositions, resulting in semantic entanglement. (2) Collision-signal heterogeneity: collisions are not uniformly harmful. Some reflect genuine conflicts between semantically unrelated items, while others stem from benign redundancy or systematic data effects. To address these challenges, we propose Qualification-Aware Semantic ID Learning (QuaSID), an end-to-end framework that learns collision-qualified SIDs by selectively repelling qualified conflict pairs and scaling the repulsion strength by collision severity. QuaSID consists of two mechanisms: Hamming-guided Margin Repulsion, which translates low-Hamming SID overlaps into explicit, severity-scaled geometric constraints on the encoder space; and Conflict-Aware Valid Pair Masking, which masks protocol-induced benign overlaps to denoise repulsion supervision. In addition, QuaSID incorporates a dual-tower contrastive objective to inject collaborative signals into tokenization. Experiments on public benchmarks and industrial data validate QuaSID. On public datasets, QuaSID consistently outperforms strong baselines, improving top-K ranking quality by 5.9% over the best baseline while increasing SID composition diversity. In an online A/B test on Kuaishou e-commerce with a 5% traffic split, QuaSID increases ranking GMV-S2 by 2.38% and improves completed orders on cold-start retrieval by up to 6.42%. Finally, we show that the proposed repulsion loss is plug-and-play and enhances a range of SID learning frameworks across datasets.

cs.IR↗

arXiv2025

Uncovering multi-technology convergence patterns with hypergraphs: Evolution and prediction using patent data

Technology convergence integrates distinct domains to create novel combinations, driving radical innovation that reshapes markets and industries. However, most approaches rely on pairwise networks that cannot capture multi-technology interactions and suffer scale biases from heterogeneous patenting activity. To overcome these gaps, we propose a hypergraph-based framework that directly models multi-technology convergence and identifies statistically significant convergence via a probabilistic null model. Using four decades of USPTO patent data (1984-2023), we construct two hypergraphs: a co-classification-based hypergraph representing explicit inventive convergence and a co-citation-based hypergraph capturing implicit knowledge-flow convergence. Evolution analysis on both hypergraph types reveals a sustained growth in multi-technology and cross-domain convergence, with a marked shift from chemistry-led to computing-led convergence patterns. Building on these insights, we formulate the forecasting of technology convergence as a hyperedge prediction task. We implement random forest classifiers trained on two complementary feature sets derived from both hypergraphs: similarity features capturing structural and semantic similarities among technologies, and intrinsic features representing inherent attributes of technologies. Predictive results demonstrate that co-citation-based features exhibit stronger predictive power than co-classification-based ones, with their combination achieving optimal performance. Explainable AI analyses identify global knowledge-flow reachability and semantic similarity as dominant drivers of convergence, while intrinsic citation and economic values exhibit contrasting associations with convergence probability. This framework bridges evolution analysis and predictive modeling of multi-technology convergence, offering actionable insights for innovation strategy.

physics.soc-ph↗

arXiv2024

Bridging the User-side Knowledge Gap in Knowledge-aware Recommendations with Large Language Models

In recent years, knowledge graphs have been integrated into recommender systems as item-side auxiliary information, enhancing recommendation accuracy. However, constructing and integrating structural user-side knowledge remains a significant challenge due to the improper granularity and inherent scarcity of user-side features. Recent advancements in Large Language Models (LLMs) offer the potential to bridge this gap by leveraging their human behavior understanding and extensive real-world knowledge. Nevertheless, integrating LLM-generated information into recommender systems presents challenges, including the risk of noisy information and the need for additional knowledge transfer. In this paper, we propose an LLM-based user-side knowledge inference method alongside a carefully designed recommendation framework to address these challenges. Our approach employs LLMs to infer user interests based on historical behaviors, integrating this user-side information with item-side and collaborative data to construct a hybrid structure: the Collaborative Interest Knowledge Graph (CIKG). Furthermore, we propose a CIKG-based recommendation framework that includes a user interest reconstruction module and a cross-domain contrastive learning module to mitigate potential noise and facilitate knowledge transfer. We conduct extensive experiments on three real-world datasets to validate the effectiveness of our method. Our approach achieves state-of-the-art performance compared to competitive baselines, particularly for users with sparse interactions.

cs.IR↗

arXiv2024

Association between centrality and flourishing trait: analyzing student co-occurrence networks drawn from dining activities

Comprehending the association between social capabilities and individual psychological traits is paramount for educational administrators. Presently, many studies heavily depend on online questionnaires and self-reported data, while analysis of the connection between offline social networks and mental health status remains scarce. By leveraging a public dataset encompassing on-campus dining activities over 21 weeks, we establish student co-occurrence networks and closely observe the changes in network topology over time. Empirical analysis shows that the node centralities of the student co-occurrence networks exhibit significantly positive correlation with the enhancement of the flourishing trait within the field of mental well-being. Our findings offer potential guidance for assisting students in maintaining a positive mental health status.

cs.SI↗

arXiv2020

Phase diagrams of interacting spreading dynamics in complex networks

Epidemic spreading processes in the real world can interact with each other in a cooperative, competitive, or asymmetric way, requiring a description based on coevolution dynamics. Rich phenomena such as discontinuous outbreak transitions and hystereses can arise, but a full picture of these behaviors in the parameter space is lacking. We develop a theory for interacting spreading dynamics on complex networks through spectral dimension reduction. In particular, we derive from the microscopic quenched mean-field equations a two-dimensional system in terms of the macroscopic variables, which enables a full phase diagram to be determined analytically. The diagram predicts critical phenomena that were known previously but only numerically, such as the interplay between discontinuous transition and hysteresis as well as the emergence and role of tricritical points.

physics.soc-ph↗

arXiv2020

Effects of heterogeneous self-protection awareness on resource-epidemic coevolution dynamics

Recent studies have demonstrated that the allocation of individual resources has a significant influence on the dynamics of epidemic spreading. In the real scenario, individuals have a different level of awareness for self-protection when facing the outbreak of an epidemic. To investigate the effects of the heterogeneous self-awareness distribution on the epidemic dynamics, we propose a resource-epidemic coevolution model in this paper. We first study the effects of the heterogeneous distributions of node degree and self-awareness on the epidemic dynamics on artificial networks. Through extensive simulations, we find that the heterogeneity of self-awareness distribution suppresses the outbreak of an epidemic, and the heterogeneity of degree distribution enhances the epidemic spreading. Next, we study how the correlation between node degree and self-awareness affects the epidemic dynamics. The results reveal that when the correlation is positive, the heterogeneity of self-awareness restrains the epidemic spreading. While, when there is a significant negative correlation, strong heterogeneous or strong homogeneous distribution of the self-awareness is not conducive for disease suppression. We find an optimal heterogeneity of self-awareness, at which the disease can be suppressed to the most extent. Further research shows that the epidemic threshold increases monotonously when the correlation changes from most negative to most positive, and a critical value of the correlation coefficient is found. When the coefficient is below the critical value, an optimal heterogeneity of self-awareness exists; otherwise, the epidemic threshold decreases monotonously with the decline of the self-awareness heterogeneity. At last, we verify the results on four typical real-world networks and find that the results on the real-world networks are consistent with those on the artificial network.

physics.soc-ph↗

arXiv2019

Optimizing spreading dynamics in interconnected networks

Adding edges between layers of interconnected networks is an important way to optimize the spreading dynamics. While previous studies mostly focus on the case of adding a single edge, the theoretical optimal strategy for adding multiple edges still need to be studied. In this study, based on the susceptible-infected-susceptible (SIS) model, we investigate the problem of maximizing the stationary spreading prevalence in interconnected networks. For two isolated networks, we maximize the spreading prevalence near the critical point by choosing multiple interconnecting edges. We present a theoretical analysis based on the discrete-time Markov chain approach to derive the approximate optimal strategy. The optimal inter-layer structure predicted by the strategy maximizes the spreading prevalence, meanwhile minimizes the spreading outbreak threshold for the interconnected network simultaneously. Numerical simulations on synthetic and real-world networks show that near the critical point, the proposed strategy gives better performance than connecting large degree nodes and randomly connecting.

physics.soc-ph↗

arXiv2019

Optimal interlayer structure for promoting spreading of SIS model in two-layer networks

Real-world systems, ranging from social and biological to infrastructural, can be modeled by multilayer networks. Promoting spreading dynamics in multilayer networks may significantly facilitate electronic advertising and predicting popular scientific publications. In this study, we propose a strategy for promoting the spreading dynamics of the susceptible-infected-susceptible model by adding one interconnecting edge between two isolated networks. By applying a perturbation method to the discrete Markovian chain approach, we derive an index that estimates the spreading prevalence in the interconnected network. The index can be interpreted as a variant of Katz centrality, where the adjacency matrix is replaced by a weighted matrix with weights depending on the dynamical information of the spreading process. Edges that are less infected at one end and its neighborhood but highly infected at the other will have larger weights. We verify the effectiveness of the proposed strategy on small networks by exhaustively examining all latent edges and demonstrate that performance is optimal or near-optimal. For large synthetic and real-world networks, the proposed method always outperforms other static strategies such as connecting nodes with the highest degree or eigenvector centrality.

physics.soc-ph↗

arXiv2018

Optimal resource diffusion for suppressing disease spreading in multiplex networks

Resource diffusion is an ubiquitous phenomenon, but how it impacts epidemic spreading has received little study. We propose a model that couples epidemic spreading and resource diffusion in multiplex networks. The spread of disease in a physical contact layer and the recovery of the infected nodes are both strongly dependent upon resources supplied by their counterparts in the social layer. The generation and diffusion of resources in the social layer are in turn strongly dependent upon the state of the nodes in the physical contact layer. Resources diffuse preferentially or randomly in this model. To quantify the degree of preferential diffusion, a bias parameter that controls the resource diffusion is proposed. We conduct extensive simulations and find that the preferential resource diffusion can change phase transition type of the fraction of infected nodes. When the degree of interlayer correlation is below a critical value, increasing the bias parameter changes the phase transition from double continuous to single continuous. When the degree of interlayer correlation is above a critical value, the phase transition changes from multiple continuous to first discontinuous and then to hybrid. We find hysteresis loops in the phase transition. We also find that there is an optimal resource strategy at each fixed degree of interlayer correlation where the threshold reaches a maximum and under which the disease can be maximally suppressed. In addition, the optimal controlling parameter increases as the degree of inter-layer correlation increases.

physics.soc-ph↗

arXiv2017

Dynamics of social contagions with local trend imitation

Research on social contagion dynamics has not yet including a theoretical analysis of the ubiquitous local trend imitation (LTI) characteristic. We propose a social contagion model with a tent-like adoption probability distribution to investigate the effect of this LTI characteristic on behavior spreading. We also propose a generalized edge-based compartmental theory to describe the proposed model. Through extensive numerical simulations and theoretical analyses, we find a crossover in the phase transition: when the LTI capacity is strong, the growth of the final behavior adoption size exhibits a second-order phase transition. When the LTI capacity is weak, we see a first-order phase transition. For a given behavioral information transmission probability, there is an optimal LTI capacity that maximizes the final behavior adoption size. Finally we find that the above phenomena are not qualitatively affected by the heterogeneous degree distribution. Our suggested theory agrees with the simulation results.

physics.soc-ph↗

arXiv2017

Suppressing epidemic spreading in multiplex networks with social-support

Although suppressing the spread of a disease is usually achieved by investing in public resources, in the real world only a small percentage of the population have access to government assistance when there is an outbreak, and most must rely on resources from family or friends. We study the dynamics of disease spreading in social-contact multiplex networks when the recovery of infected nodes depends on resources from healthy neighbors in the social layer. We investigate how degree heterogeneity affects the spreading dynamics. Using theoretical analysis and simulations we find that degree heterogeneity promotes disease spreading. The phase transition of the infected density is hybrid and increases smoothly from zero to a finite small value at the first invasion threshold and then suddenly jumps at the second invasion threshold. We also find a hysteresis loop in the transition of the infected density. We further investigate how an overlap in the edges between two layers affects the spreading dynamics. We find that when the amount of overlap is smaller than a critical value the phase transition is hybrid and there is a hysteresis loop, otherwise the phase transition is continuous and the hysteresis loop vanishes. In addition, the edge overlap allows an epidemic outbreak when the transmission rate is below the first invasion threshold, but suppresses any explosive transition when the transmission rate is above the first invasion threshold.

physics.soc-ph↗

arXiv2015

Predicting missing links via significant paths

Link prediction plays an important role in understanding intrinsic evolving mechanisms of networks. With the belief that the likelihood of the existence of a link between two nodes is strongly related with their similarity, many methods have been proposed to calculate node similarity based on node attributes and/or topological structures. Among a large variety of methods that take into account paths connecting the target pair of nodes, most of which neglect the heterogeneity of those paths. Our hypothesis is that a path consisting of small-degree nodes provides a strong evidence of similarity between two ends, accordingly, we propose a so-called sig- nificant path index in this Letter to leverage intermediate nodes' degrees in similarity calculation. Empirical experiments on twelve disparate real networks demonstrate that the proposed index outperforms the mainstream link prediction baselines.

physics.soc-ph↗

arXiv2014

Personalized recommendation with corrected similarity

Personalized recommendation attracts a surge of interdisciplinary researches. Especially, similarity based methods in applications of real recommendation systems achieve great success. However, the computations of similarities are overestimated or underestimated outstandingly due to the defective strategy of unidirectional similarity estimation. In this paper, we solve this drawback by leveraging mutual correction of forward and backward similarity estimations, and propose a new personalized recommendation index, i.e., corrected similarity based inference (CSI). Through extensive experiments on four benchmark datasets, the results show a greater improvement of CSI in comparison with these mainstream baselines. And the detailed analysis is presented to unveil and understand the origin of such difference between CSI and mainstream indices.

cs.IR↗

arXiv2014

Personalized recommendation against crowd's popular selection

The problem of personalized recommendation in an ocean of data attracts more and more attention recently. Most traditional researches ignore the popularity of the recommended object, which resulting in low personality and accuracy. In this Letter, we proposed a personalized recommendation method based on weighted object network, punishing the recommended object that is the crowd's popular selection, namely, Anti-popularity index(AP), which can give enhanced personality, accuracy and diversity in contrast to mainstream baselines with a low computational complexity.

cs.IR↗

arXiv2011

Unfolding Substructures of Complex Networks by Coupling Chaotic Oscillators beyond Global Synchronization Regime

In the past decade, synchronization on complex networks has attracted increasing attentions from various research disciplines. Most previous works, however, focus only on the dynamic behaviors of synchronization process in the stable region, i.e., global synchronization. In this letter, we demonstrate that synchronization process on complex networks can efficiently reveal the substructures of networks when the coupling strength of chaotic oscillators is under the lower boundary of stable region. Both analytic and numerical results show that the nodes belonging to the same component in the hierarchical network are tightly clustered according to the Euclidean distances between the state vectors of the corresponding oscillators, and different levels of hierarchy can be systematically unfolded by gradually tuning the coupling strength. When the coupling strengths exceed the upper boundary of stable region, the hierarchy of the network cannot be recognized by this approach. Extensive simulations suggest that our method may provide a powerful tool to detect the hierarchical community structure of complex systems and deep insight into the relationship between structure and dynamics of complex systems.

physics.data-an↗