Search

In recent years, graph contrastive learning (GCL) has received increasing attention in recommender systems due to its effectiveness in reducing bias caused by data sparsity. However, most existing GCL models rely on heuristic approaches and usually assume entity independence when constructing contrastive views. We argue that these methods struggle to strike a balance between semantic invariance and view hardness across the dynamic training process, both of which are critical factors in graph contrastive learning. To address the above issues, we propose a novel GCL-based recommendation framework RGCL, which effectively maintains the semantic invariance of contrastive pairs and dynamically adapts as the model capability evolves through the training process. Specifically, RGCL first introduces decision boundary-aware adversarial perturbations to constrain the exploration space of contrastive augmented views, avoiding the decrease of task-specific information. Furthermore, to incorporate global user-user and item-item collaboration relationships for guiding on the generation of hard contrastive views, we propose an adversarial-contrastive learning objective to construct a relation-aware view-generator. Besides, considering that unsupervised GCL could potentially narrower margins between data points and the decision boundary, resulting in decreased model robustness, we introduce the adversarial examples based on maximum perturbations to achieve margin maximization. We also provide theoretical analyses on the effectiveness of our designs. Through extensive experiments on five public datasets, we demonstrate the superiority of RGCL compared against twelve baseline models., Comment: KDD 2024

The lifelong user behavior sequence provides abundant information of user preference and gains impressive improvement in the recommendation task, however increases computational consumption significantly. To meet the severe latency requirement in online service, a short sub-sequence is sampled based on similarity to the target item. Unfortunately, items not in the sub-sequence are abandoned, leading to serious information loss. In this paper, we propose a new efficient paradigm to model the full lifelong sequence, which is named as \textbf{I}nteraction \textbf{F}idelity \textbf{A}ttention (\textbf{IFA}). In IFA, we input all target items in the candidate set into the model at once, and leverage linear transformer to reduce the time complexity of the cross attention between the candidate set and the sequence without any interaction information loss. We also additionally model the relationship of all target items for optimal set generation, and design loss function for better consistency of training and inference. We demonstrate the effectiveness and efficiency of our model by off-line and online experiments in the recommender system of Kuaishou., Comment: 7 pages, 2 figures

We report the detection of a long X-ray burst triggered on MJD 60171.65 from the ultra-compact binary 4U 1850$-$087 by the Monitor of All-sky X-ray Image and Neutron Star Interior Composition Explorer (NICER). We analyse the NICER data observed in between MJD 60095.19$-$60177.43, including one observation covered part of the long X-ray burst tail, i.e., $0.15-3.8$ hr after the trigger. The persistent spectra are quite similar and well described by a combination of multi-color disk blackbody, with the inner temperature of 0.5 keV, and a thermally comptonized continuum with the asymptotic power-law photon index of $\Gamma\sim2.2$, and electron temperature of $kT_{\rm e}\sim20-30$ keV. The persistent fluxes were around $3.8\times10^{-10}~{\rm erg~cm^{-2}~s^{-1}}$, corresponding to a local accretion rate of $1\%~\dot{m}_{\rm Edd}$. Part of time-resolved burst spectra show a clear deviation from the blackbody model, which can be improved by considering the enhanced persistent emission due to the Poynting-Robertson drag, or the reflected disk emission illuminated by the burst. From the burst flux during the cooling tail, we estimate the burst duration, $\tau \approx 0.78$ hr, the burst fluence, $E_\mathrm{b} \approx 4.1 \times 10^{41}$ ergs, and the ignition column depth, $y_{\rm ign}\approx 3.5\times10^{10}~{\rm g~cm^{-2}}$. We propose that the long X-ray burst is powered by unstable burning of pure helium in deep layer. Moreover, we identify significant 1 keV emission lines in the burst spectra, which may originate from the surrounding disk., Comment: 9 pages, 6 figures, to match the published version in ApJ

ChatGPT has achieved remarkable success in natural language understanding. Considering that recommendation is indeed a conversation between users and the system with items as words, which has similar underlying pattern with ChatGPT, we design a new chat framework in item index level for the recommendation task. Our novelty mainly contains three parts: model, training and inference. For the model part, we adopt Generative Pre-training Transformer (GPT) as the sequential recommendation model and design a user modular to capture personalized information. For the training part, we adopt the two-stage paradigm of ChatGPT, including pre-training and fine-tuning. In the pre-training stage, we train GPT model by auto-regression. In the fine-tuning stage, we train the model with prompts, which include both the newly-generated results from the model and the user's feedback. For the inference part, we predict several user interests as user representations in an autoregressive manner. For each interest vector, we recall several items with the highest similarity and merge the items recalled by all interest vectors into the final result. We conduct experiments with both offline public datasets and online A/B test to demonstrate the effectiveness of our proposed method.

Type I X-ray bursts in the ultracompact X-ray binary 4U 1820$-$30 are powered by the unstable thermonuclear burning of hydrogen-deficient material. We report the detection of 15 type I X-ray bursts from 4U 1820$-$30 observed by NICER in between 2017 and 2023. All these bursts occurred in the low state for the persistent flux in the range of $2.5-8\times10^{-9}~{\rm erg~s^{-1}~cm^{-2}}$ in 0.1$-$250 keV. The burst spectra during the tail can be well explained by blackbody model. However, for the first $\sim$5 s after the burst onset, the time-resolved spectra showed strong deviations from the blackbody model. The significant improvement of the fit can be obtained by taking into account of the enhanced persistent emission due to the Poynting-Robterson drag, the extra emission modelled by another blackbody component or by the reflection from the surrounding accretion disk. The reflection model provides a self-consistent and physically motivated explanation. We find that the accretion disk density changed with 0.5 s delay as response to the burst radiation, which indicates the distortion of the accretion disk during X-ray bursts. From the time-resolved spectroscopy, all bursts showed the characteristic of photospheric radius expansion (PRE). We find one superexpansion burst with the extreme photospheric radius $r_{\rm ph}>10^3$ km and blackbody temperature of $\sim 0.2$ keV, thirteen strong PRE bursts for $r_{\rm ph}>10^2$ km, and one moderate PRE burst for $r_{\rm ph}\sim55$ km., Comment: 16 pages, 13 figures, 4 tables. It matches the published version in Astronomy & Astrophysics

NICER observed two outbursts from the neutron star low-mass X-ray binary 4U~1730$-$22 in 2021 and 2022, which showed a similar spectral evolution in the hardness-intensity diagram. Seventeen type I X-ray bursts were identified in both outbursts. The X-ray burst spectra showed clear deviations from the blackbody model, firstly $\sim10$ s after onset. Adding the enhanced persistent emission due to the Poynting-Robertson drag or the reflection from the accretion disk both significantly improved the fitting results. We found that 12 out of 17 X-ray bursts showed the photospheric radius expansion (PRE) characteristic. Considering the nine PRE bursts out of ten X-ray bursts observed by Insight-HXMT, 78\% of bursts from 4U~1730--22 exhibited PRE. According to the burst rise time, the duration, the local accretion rate, and the burst fuel composition estimated from recurrence time, we propose that these PRE bursts were powered by pure helium. From the touchdown flux of PRE bursts, we estimate the source distance of $ d=7.54\pm{0.46} (X=0)$ kpc for a canonical neutron star with $M_{\rm NS}=1.4M_\odot$ and $R_{\rm NS}=10~{\rm km}$., Comment: 11 pages, 6 figures; Matches the version published in A&A

Type-I X-ray burst oscillations are powered by thermonuclear released on the neutron star (NS) surface in low mass X-ray binaries (LMXBs), where the burst oscillation frequencies are close to the NS spin rates. In this work, we report the detection of oscillation at 584.65 Hz during the cooling tail of a type-I X-ray bursts observed from the accreting NS LMXB 4U~1730--22 in 2022 March 20, by the \textit{ Neutron star Interior Composition Explorer} (\textit{NICER}) telescope. The oscillation signal showed a strong Leahy power, $P_{\rm m}\sim54.04$, around 584.65 Hz, which has single trial and multiple trials confidence levels of $7.05\sigma$ and $4.78\sigma$, respectively. The folded pulse profile of the oscillation in the 0.2--10 keV band showed a sinusoidal shape with the fraction amplitude rms of $(12.5\pm1.8)\%$. We found the oscillation frequency showed insignificant upward drifting, i.e., less than 0.3 Hz, during the cooling tail, similar as the behavior appearing in accreting millisecond X-ray pulsars (AMXP), and indicate the source could be an AMXP spinning at 1.71 ms., Comment: matches published version

Implicit feedback data is extensively explored in recommendation as it is easy to collect and generally applicable. However, predicting users' preference on implicit feedback data is a challenging task since we can only observe positive (voted) samples and unvoted samples. It is difficult to distinguish between the negative samples and unlabeled positive samples from the unvoted ones. Existing works, such as Bayesian Personalized Ranking (BPR), sample unvoted items as negative samples uniformly, therefore suffer from a critical noisy-label issue. To address this gap, we design an adaptive sampler based on noisy-label robust learning for implicit feedback data. To formulate the issue, we first introduce Bayesian Point-wise Optimization (BPO) to learn a model, e.g., Matrix Factorization (MF), by maximum likelihood estimation. We predict users' preferences with the model and learn it by maximizing likelihood of observed data labels, i.e., a user prefers her positive samples and has no interests in her unvoted samples. However, in reality, a user may have interests in some of her unvoted samples, which are indeed positive samples mislabeled as negative ones. We then consider the risk of these noisy labels, and propose a Noisy-label Robust BPO (NBPO). NBPO also maximizes the observation likelihood while connects users' preference and observed labels by the likelihood of label flipping based on the Bayes' theorem. In NBPO, a user prefers her true positive samples and shows no interests in her true negative samples, hence the optimization quality is dramatically improved. Extensive experiments on two public real-world datasets show the significant improvement of our proposed optimization methods., Comment: SIGIR 2020 paper

Data sparsity is an inherent challenge in the recommender systems, where most of the data is collected from the implicit feedbacks of users. This causes two difficulties in designing effective algorithms: first, the majority of users only have a few interactions with the system and there is no enough data for learning; second, there are no negative samples in the implicit feedbacks and it is a common practice to perform negative sampling to generate negative samples. However, this leads to a consequence that many potential positive samples are mislabeled as negative ones and data sparsity would exacerbate the mislabeling problem. To solve these difficulties, we regard the problem of recommendation on sparse implicit feedbacks as a semi-supervised learning task, and explore domain adaption to solve it. We transfer the knowledge learned from dense data to sparse data and we focus on the most challenging case -- there is no user or item overlap. In this extreme case, aligning embeddings of two datasets directly is rather sub-optimal since the two latent spaces encode very different information. As such, we adopt domain-invariant textual features as the anchor points to align the latent spaces. To align the embeddings, we extract the textual features for each user and item and feed them into a domain classifier with the embeddings of users and items. The embeddings are trained to puzzle the classifier and textual features are fixed as anchor points. By domain adaptation, the distribution pattern in the source domain is transferred to the target domain. As the target part can be supervised by domain adaptation, we abandon negative sampling in target dataset to avoid label noise. We adopt three pairs of real-world datasets to validate the effectiveness of our transfer strategy. Results show that our models outperform existing models significantly., Comment: KDD 2020 paper

\textbf{G}raph \textbf{C}onvolutional \textbf{N}etwork (\textbf{GCN}) is widely used in graph data learning tasks such as recommendation. However, when facing a large graph, the graph convolution is very computationally expensive thus is simplified in all existing GCNs, yet is seriously impaired due to the oversimplification. To address this gap, we leverage the \textit{original graph convolution} in GCN and propose a \textbf{L}ow-pass \textbf{C}ollaborative \textbf{F}ilter (\textbf{LCF}) to make it applicable to the large graph. LCF is designed to remove the noise caused by exposure and quantization in the observed data, and it also reduces the complexity of graph convolution in an unscathed way. Experiments show that LCF improves the effectiveness and efficiency of graph convolution and our GCN outperforms existing GCNs significantly. Codes are available on \url{https://github.com/Wenhui-Yu/LCFN}., Comment: ICML 2020 paper

To enhance the performance of the recommender system, side information is extensively explored with various features (e.g., visual features and textual features). However, there are some demerits of side information: (1) the extra data is not always available in all recommendation tasks; (2) it is only for items, there is seldom high-level feature describing users. To address these gaps, we introduce the spectral features extracted from two hypergraph structures of the purchase records. Spectral features describe the \textit{similarity} of users/items in the graph space, which is critical for recommendation. We leverage spectral features to model the users' preference and items' properties by incorporating them into a Matrix Factorization (MF) model. In addition to modeling, we also use spectral features to optimize. Bayesian Personalized Ranking (BPR) is extensively leveraged to optimize models in implicit feedback data. However, in BPR, all missing values are regarded as negative samples equally while many of them are indeed unseen positive ones. We enrich the positive samples by calculating the similarity among users/items by the spectral features. The key ideas are: (1) similar users shall have similar preference on the same item; (2) a user shall have similar perception on similar items. Extensive experiments on two real-world datasets demonstrate the usefulness of the spectral features and the effectiveness of our spectrum-enhanced pairwise optimization. Our models outperform several state-of-the-art models significantly., Comment: 11 pages; submitted to World Wide Web Conference (WWW 2019)

Skyline, aiming at finding a Pareto optimal subset of points in a multi-dimensional dataset, has gained great interest due to its extensive use for multi-criteria analysis and decision making. The skyline consists of all points that are not dominated by any other points. It is a candidate set of the optimal solution, which depends on a specific evaluation criterion for optimum. However, conventional skyline queries, which return individual points, are inadequate in group querying case since optimal combinations are required. To address this gap, we study the skyline computation in the group level and propose efficient methods to find the Group-based skyline (G-skyline). For computing the front $l$ skyline layers, we lay out an efficient approach that does the search concurrently on each dimension and investigates each point in the subspace. After that, we present a novel structure to construct the G-skyline with a queue of combinations of the first-layer points. We further demonstrate that the G-skyline is a complete candidate set of top-$l$ solutions, which is the main superiority over previous group-based skyline definitions. However, as G-skyline is complete, it contains a large number of groups which can make it impractical. To represent the "contour" of the G-skyline, we define the Representative G-skyline (RG-skyline). Then, we propose a Group-based clustering (G-clustering) algorithm to find out RG-skyline groups. Experimental results show that our algorithms are several orders of magnitude faster than the previous work., Comment: 17 pages, submitted to Transactions on Knowledge and Data Engineering (TKDE)

Visual information plays a critical role in human decision-making process. While recent developments on visually-aware recommender systems have taken the product image into account, none of them has considered the aesthetic aspect. We argue that the aesthetic factor is very important in modeling and predicting users' preferences, especially for some fashion-related domains like clothing and jewelry. This work addresses the need of modeling aesthetic information in visually-aware recommender systems. Technically speaking, we make three key contributions in leveraging deep aesthetic features: (1) To describe the aesthetics of products, we introduce the aesthetic features extracted from product images by a deep aesthetic network. We incorporate these features into recommender system to model users' preferences in the aesthetic aspect. (2) Since in clothing recommendation, time is very important for users to make decision, we design a new tensor decomposition model for implicit feedback data. The aesthetic features are then injected to the basic tensor model to capture the temporal dynamics of aesthetic preferences (e.g., seasonal patterns). (3) We also use the aesthetic features to optimize the learning strategy on implicit feedback data. We enrich the pairwise training samples by considering the similarity among items in the visual space and graph space; the key idea is that a user may likely have similar perception on similar items. We perform extensive experiments on several real-world datasets and demonstrate the usefulness of aesthetic features and the effectiveness of our proposed methods., Comment: Accepted by VLDBJ. arXiv admin note: substantial text overlap with arXiv:1809.05822