Search for: All records

Network representations have been shown to improve performance within a variety of tasks, including classification, clustering, and link prediction. However, most models either focus on moderate-sized, homogeneous networks or require a significant amount of auxiliary input to be provided by the user. Moreover, few works have studied network representations in real-world heterogeneous social networks with ambiguous social connections and are often incomplete. In the present work, we investigate the problem of learning low-dimensional node representations in heterogeneous professional social networks (HPSNs), which are incomplete and have ambiguous social connections. We present a general heterogeneous network representation learning model called Star2Vec that learns entity and person embeddings jointly using a social connection strength-aware biased random walk combined with a node-structure expansion function. Experiments on LinkedIn's Economic Graph and publicly available snapshots of Facebook's network show that Star2Vec outperforms existing methods on members' industry and social circle classification, skill and title clustering, and member-entity link predictions. We also conducted large-scale case studies to demonstrate practical applications of the Star2Vec embeddings trained on LinkedIn's Economic Graph such as next career move, alternative career suggestions, and general entity similarity searches. 

Abstract We use a combination of experiments, numerical analysis and theory to investigate the nonlinear dynamic response of a chain of precompressed elastic beams. Our results show that this simple system offers a rich platform to study the propagation of large amplitude waves. Compression waves are strongly dispersive, whereas rarefaction pulses propagate in the form of solitons. Further, we find that the model describing our structure closely resembles those introduced to characterize the dynamics of several molecular chains and macromolecular crystals, suggesting that our macroscopic system can provide insights into the effect of nonlinear vibrations on molecular mechanisms. 

Abstract We combine experimental, numerical, and analytical tools to design highly nonlinear mechanical metamaterials that exhibit a new phenomenon: gaps in amplitude for elastic vector solitons (i.e., ranges in amplitude where elastic soliton propagation is forbidden). Such gaps are fundamentally different from the spectral gaps in frequency typically observed in linear phononic crystals and acoustic metamaterials and are induced by the lack of strong coupling between the two polarizations of the vector soliton. We show that the amplitude gaps are a robust feature of our system and that their width can be controlled both by varying the structural properties of the units and by breaking the symmetry in the underlying geometry. Moreover, we demonstrate that amplitude gaps provide new opportunities to manipulate highly nonlinear elastic pulses, as demonstrated by the designed soliton splitters and diodes. 

Abstract Designing stable Li metal and supporting solid structures (SSS) is of fundamental importance in rechargeable Li‐metal batteries. Yet, the stripping kinetics of Li metal and its mechanical effect on the supporting solids (including solid electrolyte interface) remain mysterious to date. Here, through nanoscale in situ observations of a solid‐state Li‐metal battery in an electron microscope, two distinct cavitation‐mediated Li stripping modes controlled by the ratio of the SSS thickness (t) to the Li deposit's radius (r) are discovered. A quantitative criterion is established to understand the damage tolerance of SSS on the Li‐metal stripping pathways. For mechanically unstable SSS (t/r < 0.21), the stripping proceeds via tension‐induced multisite cavitation accompanied by severe SSS buckling and necking, ultimately leading to Li “trapping” or “dead Li” formation; for mechanically stable SSS (t/r > 0.21), the Li metal undergoes nearly planar stripping from the root via single cavitation, showing negligible buckling. This work proves the existence of an electronically conductive precursor film coated on the interior of solid electrolytes that however can be mechanically damaged, and it is of potential importance to the design of delicate Li‐metal supporting structures to high‐performance solid‐state Li‐metal batteries. 

With increased globalization and labor mobility, human resource reallocation across firms, industries and regions has become the new norm in labor markets. The emergence of massive digital traces of such mobility offers a unique opportunity to understand labor mobility at an unprecedented scale and granularity. While most studies on labor mobility have largely focused on characterizing macro-level (e.g., region or company) or micro-level (e.g., employee) patterns, the problem of how to accurately predict an employee's next career move (which company with what job title) receives little attention. This paper presents the first study of large-scale experiments for predicting next career moves. We focus on two sources of predictive signals: profile context matching and career path mining and propose a contextual LSTM model, NEMO, to simultaneously capture signals from both sources by jointly learning latent representations for different types of entities (e.g., employees, skills, companies) that appear in different sources. In particular, NEMO generates the contextual representation by aggregating all the profile information and explores the dependencies in the career paths through the Long Short-Term Memory (LSTM) networks. Extensive experiments on a large, real-world LinkedIn dataset show that NEMO significantly outperforms strong baselines and also reveal interesting insights in micro-level labor mobility.