Recommendation models based on deep learning are fragile when facing adversarial examples (AE). Adversarial training (AT) is the existing mainstream method to promote the adversarial robustness of recommendation models. However, these AT methods often have two drawbacks. First, they may be ineffective due to the ubiquitous sparsity of interaction data. Second, point-wise perturbation used by these AT methods leads to suboptimal adversarial robustness, because not all examples are equally susceptible to such perturbations. To overcome these issues, we propose a novel method called Cross-domain Distributional Adversarial Training (CDAT) which utilizes a richer auxiliary domain to improve the adversarial robustness of a sparse target domain. CDAT comprises a Domain adversarial network (Dan) and a Cross-domain adversarial example generative network (Cdan). Dan learns a domain-invariant preference distribution which is obtained by aligning user embeddings from two domains and paves the way to leverage the knowledge from another domain for the target domain. Then, by adversarially perturbing the domain-invariant preference distribution under the guidance of a discriminator, Cdan captures an aggressive and imperceptible AE distribution. In this way, CDAT can transfer distributional adversarial robustness from the auxiliary domain to the target domain. The extensive experiments conducted on real datasets demonstrate the remarkable superiority of the proposed CDAT in improving the adversarial robustness of the sparse domain. The codes and datasets are available on https://github.com/HymanLoveGIN/CDAT.

Personalized fairness in recommendations has been attracting increasing attention from researchers. The existing works often treat a fairness requirement, represented as a collection of sensitive attributes, as a hyper-parameter, and pursue extreme fairness by completely removing information of sensitive attributes from the learned fair embedding, which suffer from two challenges: huge training cost incurred by the explosion of attribute combinations, and the suboptimal trade-off between fairness and accuracy. In this paper, we propose a novel Adaptive Fair Representation Learning (AFRL) model, which achieves a real personalized fairness due to its advantage of training only one model to adaptively serve different fairness requirements during inference phase. Particularly, AFRL treats fairness requirements as inputs and can learn an attribute-specific embedding for each attribute from the unfair user embedding, which endows AFRL with the adaptability during inference phase to determine the non-sensitive attributes under the guidance of the user's unique fairness requirement. To achieve a better trade-off between fairness and accuracy in recommendations, AFRL conducts a novel Information Alignment to exactly preserve discriminative information of non-sensitive attributes and incorporate a debiased collaborative embedding into the fair embedding to capture attribute-independent collaborative signals, without loss of fairness. Finally, the extensive experiments conducted on real datasets together with the sound theoretical analysis demonstrate the superiority of AFRL.

The cold-start problem is a long-standing challenge in recommender systems. As a promising solution, content-based generative models usually project a cold-start item’s content onto a warm-start item embedding to capture collaborative signals from item content so that collaborative filtering can be applied. However, since the training of the cold-start recommendation models is conducted on warm datasets, the existent methods face the issue that the collaborative embeddings of items will be blurred, which significantly degenerates the performance of cold-start item recommendation. To address this issue, we propose a novel model called Contrastive Collaborative Filtering for Cold-start item Recommendation (CCFCRec), which capitalizes on the co-occurrence collaborative signals in warm training data to alleviate the issue of blurry collaborative embeddings for cold-start item recommendation. In particular, we devise a contrastive collaborative filtering (CF) framework, consisting of a content CF module and a co-occurrence CF module to generate the content-based collaborative embedding and the co-occurrence collaborative embedding for a training item, respectively. During the joint training of the two CF modules, we apply a contrastive learning between the two collaborative embeddings, by which the knowledge about the co-occurrence signals can be indirectly transferred to the content CF module, so that the blurry collaborative embeddings can be rectified implicitly by the memorized co-occurrence collaborative signals during the applying phase. Together with the sound theoretical analysis, the extensive experiments conducted on real datasets demonstrate the superiority of the proposed model. The codes and datasets are available on https://github.com/zzhin/CCFCRec.

Traditional robust recommendation methods view atypical user-item interactions as noise and aim to reduce their impact with some kind of noise filtering technique, which often suffers from two challenges. First, in real world, atypical interactions may signal users’ temporary interest different from their general preference. Therefore, simply filtering out the atypical interactions as noise may be inappropriate and degrade the personalization of recommendations. Second, it is hard to acquire the temporary interest since there are no explicit supervision signals to indicate whether an interaction is atypical or not. To address this challenges, we propose a novel model called Temporary Interest Aware Recommendation (TIARec), which can distinguish atypical interactions from normal ones without supervision and capture the temporary interest as well as the general preference of users. Particularly, we propose a reinforcement learning framework containing a recommender agent and an auxiliary classifier agent, which are jointly trained with the objective of maximizing the cumulative return of the recommendations made by the recommender agent. During the joint training process, the classifier agent can judge whether the interaction with an item recommended by the recommender agent is atypical, and the knowledge about learning temporary interest from atypical interactions can be transferred to the recommender agent, which makes the recommender agent able to alone make recommendations that balance the general preference and temporary interest of users. At last, the experiments conducted on real world datasets verify the effectiveness of TIARec.

Robust recommendation aims at capturing true preference of users from noisy data, for which there are two lines of methods have been proposed. One is based on noise injection, and the other is to adopt the generative model Variational Auto-encoder (VAE). However, the existing works still face two challenges. First, the noise injection based methods often draw the noise from a fixed noise distribution given in advance, while in real world, the noise distributions of different users and items may differ from each other due to personal behaviors and item usage patterns. Second, the VAE based models are not expressive enough to capture the true preference since VAE often yields an embedding space of a single modal, while in real world, user-item interactions usually exhibit multi-modality on user preference distribution. In this paper, we propose a novel model called Dual Adversarial Variational Embedding (DAVE) for robust recommendation, which can provide personalized noise reduction for different users and items, and capture the multi-modality of the embedding space, by combining the advantages of VAE and adversarial training between the introduced auxiliary discriminators and the variational inference networks. The extensive experiments conducted on real datasets verify the effectiveness of DAVE on robust recommendation.

Although a variety of methods have been proposed for sequential recommendation, it is still far from being well solved partly due to two challenges. First, the existing methods often lack the simultaneous consideration of the global stability and local fluctuation of user preference, which might degrade the learning of a user's current preference. Second, the existing methods often use a scalar based weighting schema to fuse the long-term and short-term preferences, which is too coarse to learn an expressive embedding of current preference. To address the two challenges, we propose a novel model called Time Lag aware Sequential Recommendation (TLSRec), which integrates a hierarchical modeling of user preference and a time lag sensitive fine-grained fusion of the long-term and short-term preferences. TLSRec employs a hierarchical self-attention network to learn users' preference at both global and local time scales, and a neural time gate to adaptively regulate the contributions of the long-term and short-term preferences for the learning of a user's current preference at the aspect level and based on the lag between the current time and the time of the last behavior of a user. The extensive experiments conducted on real datasets verify the effectiveness of TLSRec.

Conversational recommender systems model user dynamic preferences and recommend items based on multi-turn interactions. Though the conversational recommender system has achieved good performance, it has two limitations. On the one hand, researchers usually random select an anchor item from user's historical interactions to simulate the interaction with the real user, but some items in the historical interactions do not fit the user realistic preferences (item noise). On the other hand, it pays too much attention to user dynamic preferences, but nurses some static preferences that are difficult to change over a short period. In fact, when there is no explicit attribute preference in user's conversation, the user static preferences can also be used to make recommendations. To address the aforementioned issues, a novel method that combines graph path reasoning with multi-turn conversation is proposed, called Graph Path reasoning for conversational Recommendation (GPR). In GPR, a soft-clustering is designed to classify items and then set operations are utilized to filter the noise in the user's historical interactions. To capture user dynamic preferences and take account of the user inherent static preferences, GPR asks questions about attributes in the attribute-level reasoning and asks whether the items fit user static preferences in the item-level reasoning on a heterogeneous graph. In the multi-turn of two-level graph path reasoning, a reinforcement learning is used to obtain the optimal path and accurately recommend items to users. Extensive experiments conducted on two benchmark datasets verify that GPR can significantly improve recommendation performance and reduce the turn of path reasoning.

Cross-domain recommendation (CDR) has been attracting increasing attention of researchers for its ability to alleviate the data sparsity problem in recommender systems. However, the existing single-target or dual-target CDR methods often suffer from two drawbacks, the assumption of at least one rich domain and the heavy dependence on domain-invariant preference, which are impractical in real world where sparsity is ubiquitous and might degrade the user preference learning. To overcome these issues, we propose a Multi-Sparse-Domain Collaborative Recommendation (MSDCR) model for multi-target cross-domain recommendation. Unlike traditional CDR methods, MSDCR treats the multiple relevant domains as all sparse and can simultaneously improve the recommendation performance in each domain. We propose a Multi-Domain Separation Network (MDSN) and a Gated Aspect Preference Enhancement (GAPE) module for MSDCR to enhance a user's domain-specific aspect preferences in a domain by transferring the complementary aspect preferences in other domains, during which the uniqueness of the domain-specific preference can be preserved through the adversarial training offered by MDSN and the complementarity can be adaptively determined by GAPE. Meanwhile, we propose a Multi-Domain Adaptation Network (MDAN) for MSDCR to capture a user's domain-invariant aspect preference. With the integration of the enhanced domain-specific aspect preference and the domain-invariant aspect preference, MSDCR can reach a comprehensive understanding of a user's preference in each sparse domain. At last, the extensive experiments conducted on real datasets demonstrate the remarkable superiority of MSDCR over the state-of-the-art single-domain recommendation models and CDR models.

Most existing random walk based network embedding methods often follow only one of two principles, homophily or structural equivalence. In real world networks, however, nodes exhibit a mixture of homophily and structural equivalence, which requires adaptive network embedding that can adaptively preserve both homophily and structural equivalence for different nodes in different down-stream analysis tasks. In this paper, we propose a novel method called Meta-Learning based Adaptive Network Embedding (MLANE), which can learn adaptive sampling strategy for different nodes in different tasks by incorporating sampling strategy learning with embedding learning into one optimization problem that can be solved via an end-to-end meta-learning framework. In extensive experiments on real datasets, MLANE shows significant performance improvements over the baselines. The source code of MLANE and the datasets used in experiments and all the hyperparameter settings for baselines are available at https://github.com/7733com/MLANE.

Explainable recommendation is far from being well solved partly due to three challenges. The first is the personalization of preference learning, which requires that different items/users have different contributions to the learning of user preference or item quality. The second one is dynamic explanation, which is crucial for the timeliness of recommendation explanations. The last one is the granularity of explanations. In practice, aspect-level explanations are more persuasive than item-level or user-level ones. In this paper, to address these challenges simultaneously, we propose a novel model called Hybrid Deep Embedding (HDE) for aspect-based explainable recommendations, which can make recommendations with dynamic aspect-level explanations. The main idea of HDE is to learn the dynamic embeddings of users and items for rating prediction and the dynamic latent aspect preference/quality vectors for the generation of aspect-level explanations, through fusion of the dynamic implicit feedbacks extracted from reviews and the attentive user-item interactions. Particularly, as the aspect preference/quality of users/items is learned automatically, HDE is able to capture the impact of aspects that are not mentioned in reviews of a user or an item. The extensive experiments conducted on real datasets verify the recommending performance and explainability of HDE. The source code of our work is available at https://github.com/lola63/HDE-Python.

The purpose of diffusion history inference is to reconstruct the missing traces of information diffusion according to incomplete observations. Existing methods, however, often focus only on single diffusion trace, while in a real-world social network, there often coexist multiple information diffusions. In this paper, we propose a novel approach called Collaborative Inference Model (CIM) for the problem of the inference of coexisting information diffusions. CIM can holistically model multiple information diffusions without any prior assumption of diffusion models, and collaboratively infer the histories of the coexisting information diffusions via low-rank approximation with a fusion of heterogeneous constraints generated from additional data sources. We also propose an optimized algorithm called Time Window based Parallel Decomposition Algorithm (TWPDA) to speed up the inference without compromise on the accuracy. Extensive experiments are conducted on real-world datasets to verify the effectiveness and efficiency of CIM and TWPDA.

In this paper, we investigate the problem of reconstructing hidden trajectories from a collective of separate spatial-temporal points without ID information, given the number of hidden trajectories. The challenge is three-fold: lack of meaningful features, data sparsity, and missing trajectory links. We propose a novel approach called Hidden Trajectory Reconstruction (HTR). From an information-theoretic perspective, we devise five novel temporal features and combine them into an Latent Spatial-Temporal Feature Vector (LSTFV) to characterize the dynamics of a single spatial-temporal point. The proposed features have the potential of distinguishing spatial-temporal points between trajectories. To overcome the data sparsity, we assemble the LSTFVs to a sparse Temporal Feature Tensor (TF-Tensor) and propose an algorithm called Parallel Iterative Collaborative Approximation of Sparse Tensor (PICAST). PICAST approximates the TF-Tensor by decomposing it into a tensor product of a low-rank core identity tensor and three dense factor matrices with a divide-and-conquer strategy. To achieve a dense approximate tensor with good accuracy and efficiency, PICAST minimizes a sparsity-measure and fuses an additional matrix of static geographical region features. To recover the missing trajectory links, we propose a mapping, Cross-Temporal Connectivity Preserving Transformation (CTCPT), to map the LSTFVs of the separate spatial-temporal points to an intrinsic space called Cross-Temporal Connectivity Preserving Space (CTCPS). CTCPT uses Cross-Temporal Connectivity (CTC) to evaluate whether two spatial-temporal points belong to the same trajectory and if they do, how strong the connectivity between them is. Due to the CTCPT, the hidden trajectories can be reconstructed from clusters generated in CTCPS by a clustering algorithm. At last, the extensive experiments conducted on synthetic datasets and real datasets verify the effectiveness and efficiency of our algorithms.

Recently, considerable attention has been devoted to the prediction problems arising from heterogeneous information networks. In this paper, we present a new prediction task, Neighbor Distribution Prediction (NDP), which aims at predicting the distribution of the labels on neighbors of a given node and is valuable for many different applications in heterogeneous information networks. The challenges of NDP mainly come from three aspects: the infinity of the state space of a neighbor distribution, the sparsity of available data, and how to fairly evaluate the predictions. To address these challenges, we first propose an Evolution Factor Model (EFM) for NDP, which utilizes two new structures proposed in this paper, i.e. Neighbor Distribution Vector (NDV) to represent the state of a given node's neighbors, and Neighbor Label Evolution Matrix (NLEM) to capture the dynamics of a neighbor distribution, respectively. We further propose a learning algorithm for Evolution Factor Model. To overcome the problem of data sparsity, the learning algorithm first clusters all the nodes and learns an NLEM for each cluster instead of for each node. For fairly evaluating the predicting results, we propose a new metric: Virtual Accuracy (VA), which takes into consideration both the absolute accuracy and the predictability of a node. Extensive experiments conducted on three real datasets from different domains validate the effectiveness of our proposed model EFM and metric VA.

Predicting both the time and the location of human movements is valuable but challenging for a variety of applications. To address this problem, we propose an approach considering both the periodicity and the sociality of human movements. We first define a new concept, Social Spatial-Temporal Event (SSTE), to represent social interactions among people. For the time prediction, we characterise the temporal dynamics of SSTEs with an ARMA (AutoRegressive Moving Average) model. To dynamically capture the SSTE kinetics, we propose a Kalman Filter based learning algorithm to learn and incrementally update the ARMA model as a new observation becomes available. For the location prediction, we propose a ranking model the periodicity and the sociality of human movements are simultaneously taken into consideration for improving the prediction accuracy. Extensive experiments conducted on real data sets validate our proposed approach.

Intervention analysis is the common method to reveal relationships between objects in human as well as biological society. Data mining research community is just starting to pay attention to intervention analysis. As the traditional association rules are not successful at measuring intervention, this paper tries to mining intervention rules from time series data of sub-complex system. The main contributions of this study include: (1) introduces a new concept of intervention rules. It can quantitatively detect at what scale, how intensive and how long the intervention does make sense; (2) conduct wavelet transform on time series. The decomposition scale can denotes the intervention scale; (3) proposes a new concept named directional correlation to measure intervention intensity; (4) calculates the intervention intensity between time series data with different time delays. The time delay can uncover after how long the intervention takes place; (5) conducts experiments on real datasets. The results show that intervention rules do exist at different decomposition scale of the original data. Moreover, the number of rules discovered by multi-scale analyzing methods is always 4 or 5 times more than those found by single scale methods. And the stability of rules discovered by the improved multi-scale mining algorithm MMIR* is always 100%, while the stability of rules discovered by the basic multi-scale mining algorithm MMIR fluctuate around 70%.