Search

Recommendation with implicit feedback such as “bought” in e-commerce sites and “like” in online-to-offline services is an important problem because of the abundance of users’ online behaviors. Previous works for modeling users’ implicit feedback mainly focus on designing some predefined similarity, learned similarity, or hybrid similarity between two users (or two items). However, these similarities either can not capture both the local and global relations among users (or items) simultaneously, or they are not easy to be tuned due to the tradeoff parameter in the linear hybridization. Moreover, information from one single group, either a user group or an item group, may not be sufficient for modeling users’ preferences well. To this end, in this paper, we first propose a heterogeneous similarity to capture rich correlations among users and items via a concise integration, and then design a novel recommendation algorithm, i.e., pointwise factored heterogeneous similarity model (Poi-FHSM), in which we fully exploit the proposed similarity in a flexible pointwise preference learning paradigm. Finally, we conduct extensive empirical studies on five real-world datasets, and find that our Poi-FHSM performs better than those based on the predefined, learned and hybrid similarity, as well as the state-of-the-art deep learning-based methods, showcasing the effectiveness of our heterogeneous similarity and recommendation algorithm.

Background According to the guidelines, PD-L1 expression is a critical indicator for guiding immunotherapy application. According to certain studies, regardless of PD-L1 expression, immunotherapy could be advantageous for individuals with gastric cancer. Therefore, new scoring systems or biomarkers are required to enhance treatment strategies. Methods Mass spectrometry and machine learning were used to search for strongly related PD-L1 genes, and the NMF approach was then used to separate gastric cancer patients into two categories. Differentially expressed genes (DEGs) between the two subtypes identified in this investigation were utilized to develop the UBscore predictive model, which was verified by the Gene Expression Omnibus (GEO) database. Coimmunoprecipitation, protein expression, and natural killing (NK) cell coculture experiments were conducted to validate the findings. Results A total of 123 proteins were identified as PD-L1 interactors that are substantially enriched in the proteasome complex at the mRNA level. Using random forest, 30 UPS genes were discovered in the GSE66229 cohort, and ANAPC7 was experimentally verified as one of 123 PD-L1 interactors. Depending on the expression of PD-L1 and ANAPC7, patients were separated into two subgroups with vastly distinct immune infiltration. Low UBscore was related to increased tumor mutation burden (TMB) and microsatellite instability-high (MSI-H). In addition, chemotherapy medications were more effective in individuals with a low UBscore. Finally, we discovered that ANAPC7 might lead to the incidence of immunological escape when cocultured with NK-92 cells. Conclusion According to our analysis of the PD-L1-related signature in GC, the UBscore played a crucial role in prognosis and had a strong relationship with TMB, MSI, and chemotherapeutic drug sensitivity. This research lays the groundwork for improving GC patient prognosis and treatment response.

Workplace accidents are of great concern in the construction industry. Most of those accidents are caused by unsafe behavior in the workplace. Many previous studies have analyzed the causes of workers’ unsafe behaviors, but few have investigated workers’ feelings of insecurity from the perspective of systematic psychological theory. This study developed an attitude–behavior–intervention feedback loop mechanism of construction workers and used the dual-attitude theory to explain the occurrence mechanisms of unsafe behavior. Using this mechanism, an active-intervention system-dynamics model and a passive-intervention system-dynamics model were designed and simulated. The coefficient of the system dynamics equation in the simulation model involved meta-analysis to combine the correlation coefficients of existing studies, which increased the sample size and improved the statistical test efficiency. The results show that an implicit safety attitude has a more significant impact on safety behavior, and the effect of an active intervention is stronger than that of a passive intervention. Based on these results, this paper presents some feasible suggestions to reduce the probability of unsafe worker behaviors occurring.

Recommender systems play an important role in providing personalized services for users in the context of information overload. Generally, users’ feedback toward items often contain the most significant information reflecting their preferences, which enables accurate personalized recommendation. In real applications, users’ feedback are usually heterogeneous (rather than homogeneous) such as purchases and examinations in e-commerce, which reflects users’ preferences in different degrees. Effective modeling of such heterogeneous one-class feedback is challenging compared with that of homogeneous feedback of ratings. As a response, heterogeneous one-class collaborative filtering (HOCCF) is proposed, which often converts the heterogeneous feedback into two parts (i.e., target feedback and auxiliary feedback), aiming to care more about the target feedback (e.g., purchases ) with the assistance of the auxiliary feedback (e.g., examinations ). In this survey, we provide an overview of the representative HOCCF methods from the perspective of factorization-based methods, transfer learning-based methods, and deep learning-based methods. First, we review the factorization-based methods according to different strategies. Second, we describe the transfer learning-based methods with different knowledge sharing manners. Third, we discuss the deep learning-based methods according to the neural architectures. Moreover, we include some important example applications, describe the empirical studies, and discuss some promising future directions.

Numerical modeling and experimental measurements are developing in parallel, for a long period there is a resolution gap in between, making it hard to conduct a direct comparison. In this work we numerically studied the diffusive ionization wave and fast ionization wave discharge experiments with recently published E--FISH data using a classical fluid model. The pressure--E/N range of the drift diffusion approximation and the pressure-grid range of the local field/mean energy approximation for the fluid model are proposed. The three--terms Helmholtz photo--ionization model is generalized with parameters given for N2, O2, CO2 and air. The capabilities of the classical fluid method in modeling the inception, propagation and channel breakdown stages are studied. The calculated electric field evolution of the ionization are compared with E--FISH measurements in discharge development stage and gap-closing stage, the influence of the electrode shape and predefined electron density on the streamer morphology, the long existing inception problem of the ionization waves are discussed in detail. Within the application range of the classical fluid model, a good agreement can be achieved between calculation and measurements.

In this paper, we study a recent and important recommendation problem called heterogeneous one-class collaborative filtering (HOCCF), where we have two different types of one-class feedback, i.e., a set of browses and a set of purchases, for preference learning. Previous methods exploit the browses and purchases by extending some existing methods on modeling homogeneous one-class feedback in an integrative or sequential manner. However, an integrative method may be of high complexity in model training due to the expanded prediction rule, and a sequential method is of inconvenience in deployment because more than one parametric models have to be maintained. In this paper, we convert the HOCCF problem to an adaptive transfer learning task, where we first model browses via a factorization model from a perspective of the role of browser, and then adaptively refine the learned model parameters via purchases from a perspective of the dependent role of purchaser. Based on this conversion, we design a novel solution called role-based adaptive factorization (ROAF), and then derive two specific variants with pairwise preference learning and pointwise preference learning. Finally, we conduct extensive empirical studies on two large datasets, and find that our ROAF is a very promising solution in terms of recommendation accuracy, besides its convenience of one single parametric model in deployment.

The fast ionization wave (FIW) discharges in pure nitrogen in a capillary tube at 27 mbar, initiated by positive polarity, high-voltage nanosecond pulses, are numerically studied by coupled two-dimensional plasma fluid modeling. The 2D fluid code based on local mean energy approximation is validated and used, an extended three-exponential Helmholtz photoionization model is proposed for pure nitrogen. The development and structure of the nitrogen FIW is analyzed, the electric field and current are calculated compared with experimental measurements. The evolution of radial distribution of electrons and N2(C 3Π u ) during the FIW development stage and in the afterglow are analyzed, the radial profiles of electron density show a ‘hollow’ structure in the FIW development stage, the temporal-spatial evolution of N2(C 3Π u ) is dominated by the competition between the pooling reaction of N 2 ( A 3 Σ u + ) and the quenching by electrons. The role of photoionization on the nitrogen FIW radial morphology is discussed, the equivalent background electron density of photoionization in nitrogen discharge is suggested to be (4–6) × 1013 m−3. Spatial distribution of specific energy deposition (SED), fast gas heating (FGH) energy and temperature rise are obtained, heating efficiency varies with electric field E/N and SED and tends to be 10% at high SED. The dominating reactions responsible for FGH and their fractional contribution in space and time are analyzed, in the near axis region, pooling reactions of N 2 ( A 3 Σ u + ) and N2(B 3Π g ) contribute up to 80% FGH energy, electron impact dissociation of molecular nitrogen contributes about 10%, while e- N 2 + recombination and quenching of N(2 D) atoms by N2 molecules contribute rest.

The streamer-to-spark transition in a point-to-plane configuration in atmospheric pressure air is studied using a 2D–0D combined approach. A validated fluid code is used and improved to model the spark stage. 2D modeling of discharges at three different temperatures, 300, 600 and 800 K, are conducted; the spark transition occurs when the temperature reaches 800 K in the first pulse. A conservative criteria of spark transition temperature is proposed based on analytical solution and compared with experiments. Kinetics modeling of the post discharge stage is conducted with consistent input values extracted from the 2D model. Results show that the streamer-to-spark transition can be initiated at a lower temperature (600 K) and lower field (50–75 Td), and the long-lifetime O-atoms formed in previous pulses play an important role in ‘knocking off’ the electrons from negative charged species and maintaining the electron density in the post discharge stage. The dominating processes for electron production are electron detachment reactions from O 3 − , O 2 − and O−. The ionizations from excited species only accelerate the production of electrons when the plasma is already dense.

Surface dielectric barrier discharge (SDBD) actuators driven by the Pulsed-DC voltages are analyzed. The Pulsed-DC SDBD studied in this work is equivalent to a classical SDBD driven by a tailored Fast-Rise-Slow-Decay (FRSD) voltage waveform. The plasma channel formation and charge production process in the voltage rising stage are studied at different slopes using a classical 2D fluid model, the thrust generated in the voltage decaying stage is studied based on an analytical approach taking 2D model results as the input. A thrust pulse is generated in the trailing edge of the voltage waveform and reaches maximum when the voltage decreases by approximately the value of cathode voltage fall~($\approx$600~V). The time duration of the rising and trailing edge, the decay rate and the amplitude of applied voltage are the main factors affecting the performance of the actuator. Analytical expressions are formulated for the value and time moment of peak thrust, the upper limit of thrust is also estimated. Higher voltage rising rate leads to higher charge density in the voltage rising stage thus higher thrust. Shorter voltage trailing edge, in general, results in higher value and earlier appearance of the peak thrust. The detailed profile of the trailing edge also affects the performance. Results in this work allow us to flexibly design the FRSD waveforms for an SDBD actuator according to the requirements of active flow control in different application conditions.