Sequential recommender systems, representing an advanced approach in modeling user behavior sequences, excel in capturing the dynamic evolution of user interests. However, challenges such as data sparsity, noise, and bias significantly impair the accuracy of these recommendations. This study introduces a novel counterfactual sequence generation algorithm, anchored on pivotal items. It innovates by navigating the counterfactual domain of user behavior sequences, employing a strategy to simulate alternative product selection pathways. This approach generates a variety of counterfactual sequences, thereby enriching the dataset for model training. Additionally, we present a counterfactual generation mechanism that delineates the scope of counterfactual exploration, enhancing sequence plausibility through strategic key item delineation and the incorporation of stochastic perturbations. Addressing sequence generation challenges, we propose a contrastive learning framework designed to discern and preserve the “invariance” within user interests, thus bolstering model resilience. To further refine the algorithm, we incorporate principles of alignment and uniformity within the counterfactual domain, reinforcing the algorithm's precision and overall performance. Empirical validation across four public datasets corroborates the algorithm's effectiveness, demonstrating notable advancements in recommendation accuracy and robustness.

Recommender systems not only serve users but also affect user preferences through personalized recommendations. Recent researches investigate the effects of the entire recommender system on user preferences, i.e., system-level effects, and find that recommendations may lead to problems such as echo chambers and filter bubbles. To properly alleviate the problems, it is necessary to estimate the effects of recommending a specific item on user preferences, i.e., item-level effects. For example, by understanding whether recommending an item aggravates echo chambers, we can better decide whether to recommend it or not.

Sequential recommendation has attracted a lot of attention from both academia and industry. Since item embeddings directly affect the recommendation results, their learning process is very important. However, most existing sequential models may introduce bias when updating the item embeddings. For example, in a sequence where all items are endorsed by a same celebrity, the co-occurrence of two items only indicates their similarity in terms of endorser, and is independent of the other aspects such as category and color. The existing models often update the entire item as a whole or update different aspects of the item without distinction, which fails to capture the contributions of different aspects to the co-occurrence pattern. To overcome the above limitations, we propose aspect re-distribution (ARD) to focus on updating the aspects that are important for co-occurrence. Specifically, we represent an item using several aspect embeddings with the same initial importance. We then re-calculate the importance of each aspect according to the other items in the sequence. Finally, we aggregate these aspect embeddings into a single aspect-aware embedding according to their importance. The aspect-aware embedding can be provided as input to a successor sequential model. Updates of the aspect-aware embedding are passed back to the aspect embeddings based on their importance. Therefore, different from the existing models, our method pays more attention to updating the important aspects. In our experiments, we choose self-attention networks as the successor model. The experimental results on four real-world datasets indicate that our method achieves very promising performance in comparison with seven state-of-the-art models.

This paper focuses on the multi-behavior recommendation problem, i.e., generating personalized recommendation based on multiple types of user behaviors. Methods proposed recently usually leverage the ordinal assumption, which means that users? different types of behaviors should take place in a fixed order. However, this assumption may be too strong in some scenarios. In this paper, a more general model named Multiplex Graph Neural Network (MGNN) is proposed as a remedy. MGNN tackles the multi-behavior recommendation problem from a novel perspective, i.e., the perspective of link prediction in multiplex networks. By taking advantage of both the multiplex network structure and graph representation learning techniques, MGNN learns shared embeddings and behavior-specific embeddings for users and items to model the collective effect of multiple types of behaviors. Experiments conducted on both ordinal-behavior datasets and generic-behavior datasets demonstrate the effectiveness of the proposed MGNN model.

Convolutional neural networks (CNN) are widely used on sequential data since it can capture local context dependencies and temporal order information inside sequences. Attention (ATT) mechanisms have also attracted enormous interests due to its capability of capturing the important parts of a sequence. These two neural networks can extract different features from sequences. In order to combine the advantages of CNN and ATT, we propose a convolutional attention network (CAN), which merges the structure of CNN and ATT into a single neural network and can serve as a new basic module in complex neural networks. Based on CAN, we then build a sequence encoding model with hierarchical structure, "hierarchical convolutional attention network (HiCAN)", to tackle sequence modeling problems. It can explicitly capture both the local and global context dependencies and temporal order information in sequences. Extensive experiments conducted on session-based recommendation (Recommender Systems) demonstrate that HiCAN is able to outperform state-of-the-art methods and show higher computational efficiency. Furthermore, we conduct extended experiments on text classification (Natural Language Processing). The results show that our model can also achieve competitive performance on NLP tasks.

The task of session-based recommendation aims to predict users' future interests based on anonymous historical sessions. Recent works have shown that memory models, which capture user preference from previous interaction sequence with long short-term or short-term memory, can lead to encouraging results in this problem. However, most existing memory models tend to regard each item as a memory unit, which neglect n-gram features and are insufficient to learn the user's feature-level preferences. In this paper, we aim to leverage n-gram features and model users' feature-level preferences in an explicit and effective manner. To this end, we present a memory model with multi-scale feature memory for session-based recommendation. A densely connected convolutional neural network (CNN) with short-cut path between upstream and downstream convolutional blocks is applied to build multi-scale features from item representations, and features in the same scale are combined with memory mechanism to capture users' feature-level preferences. Furthermore, attention is used to adaptively select users' multi-scale feature-level preferences for recommendation. Extensive experiments conducted on two benchmark datasets demonstrate the effectiveness of the proposed model in comparison with competitive baselines.

Utilizing knowledge graphs (KGs) to improve the performance of recommender systems has attracted increasing attention recently. Existing path-based methods rely heavily on manually designed meta-paths, while embedding-based methods focus on incorporating the knowledge graph embeddings (KGE) into recommender systems, but rarely model user-entity interactions, which can be used to enhance the performance of recommendation. To overcome the shortcomings of previous works, we propose SARC, an embedding-based model that utilizes a novel Split-And-ReCombine strategy for knowledge-based recommendation. Firstly, SARC splits the user-item-entity interactions into three 2-way interactions, i.e., the user-item, user-entity and item-entity interactions. Each of the 2-way interactions can be cast as a graph, and we use Graph Neural Networks (GNN) and KGE to model them. Secondly, SARC recombines the representation of users and items learned from the first step to generates recommendation. In order to distinguish the informative part and meaningless part of the representations, we utilize a gated fusion mechanism. The advantage of our SARC model is that through splitting, we can easily handle and make full use of the 2-way interactions, especially the user-entity interactions, and through recombining, we can extract the most useful information for recommendation. Extensive experiments on three real-world datasets demonstrate that SARC outperforms several state-of-the-art baselines.

Recommender systems are aimed at generating a personalized ranked list of items that an end user might be interested in. With the unprecedented success of deep learning in computer vision and speech recognition, recently it has been a hot topic to bridge the gap between recommender systems and deep neural network. And deep learning methods have been shown to achieve state-of-the-art on many recommendation tasks. For example, a recent model, NeuMF, first projects users and items into some shared low-dimensional latent feature space, and then employs neural nets to model the interaction between the user and item latent features to obtain state-of-the-art performance on the recommendation tasks. NeuMF assumes that the non-interacted items are inherent negative and uses negative sampling to relax this assumption. In this paper, we examine an alternative approach which does not assume that the non-interacted items are necessarily negative, just that they are less preferred than interacted items. Specifically, we develop a new classification strategy based on the widely used pairwise ranking assumption. We combine our classification strategy with the recently proposed neural collaborative filtering framework, and propose a general collaborative ranking framework called Neural Network based Collaborative Ranking (NCR). We resort to a neural network architecture to model a user's pairwise preference between items, with the belief that neural network will effectively capture the latent structure of latent factors. The experimental results on two real-world datasets show the superior performance of our models in comparison with several state-of-the-art approaches.

People consume and rate products in online shopping websites. The historical purchases of customers reflect their personal consumption habits and indicate their future shopping behaviors. Traditional preference-based recommender systems try to provide recommendations by analyzing users' feedback such as ratings and clicks. But unfortunately, most of the existing recommendation algorithms ignore the budget of the users. So they cannot avoid recommending users with products that will exceed their budgets. And they also cannot understand how the users will assign their budgets to different products. In this paper, we develop a generative model named collaborative budget-aware Poisson factorization (CBPF) to connect users' ratings and budgets. The CBPF model is intuitive and highly interpretable. We compare the proposed model with several state-of-the-art budget-unaware recommendation methods on several real-world datasets. The results show the advantage of uncovering users' budgets for recommendation.

By restricting the iterate on a nonlinear manifold, the recently proposed Riemannian optimization methods prove to be both efficient and effective in low rank tensor completion problems. However, existing methods fail to exploit the easily accessible side information, due to their format mismatch. Consequently, there is still room for improvement. To fill the gap, in this paper, a novel Riemannian model is proposed to tightly integrate the original model and the side information by overcoming their inconsistency. For this model, an efficient Riemannian conjugate gradient descent solver is devised based on a new metric that captures the curvature of the objective. Numerical experiments suggest that our method is more accurate than the state-of-the-art without compromising the efficiency.

Subspace clustering is an important unsupervised learning problem with wide applications in computer vision and data analysis. However, the state-of-the-art methods for this problem suffer from high time complexity---quadratic or cubic in $n$ (the number of data instances). In this paper we exploit a data selection algorithm to speedup computation and the robust principal component analysis to strengthen robustness. Accordingly, we devise a scalable and robust subspace clustering method which costs time only linear in $n$. We prove theoretically that under certain mild assumptions our method solves the subspace clustering problem exactly even for grossly corrupted data. Our algorithm is based on very simple ideas, yet it is the only linear time algorithm with noiseless or noisy recovery guarantee. Finally, empirical results verify our theoretical analysis.

Matrix factorization is one of the most powerful techniques in collaborative filtering, which models the (user, item) interactions behind historical explicit or implicit feedbacks. However, plain matrix factorization may not be able to uncover the structure correlations among users and items well such as communities and taxonomies. As a response, we design a novel algorithm, i.e., hierarchical group matrix factorization (HGMF), in order to explore and model the structure correlations among users and items in a principled way. Specifically, we first define four types of correlations, including (user, item), (user, item group), (user group, item) and (user group, item group); we then extend plain matrix factorization with a hierarchical group structure; finally, we design a novel clustering algorithm to mine the hidden structure correlations. In the experiments, we study the effectiveness of our HGMF for both rating prediction and item recommendation, and find that it is better than some state-of-the-art methods on several real-world data sets.

Learning users' preferences is critical to enable personalized recommendation services in various online applications such as e-commerce, entertainment and many others. In this paper, we study on how to learn users' preferences from abundant online activities, e.g., browsing and examination, which are usually called implicit feedbacks since they cannot be interpreted as users' likes or dislikes on the corresponding products directly. Pairwise preference learning algorithms are the state-of-the-art methods for this important problem, but they have two major limitations of low accuracy and low efficiency caused by noise in observed feedbacks and non-optimal learning steps in update rules. As a response, we propose a novel adaptive pairwise preference learning algorithm, which addresses the above two limitations in a single algorithm with a concise and general learning scheme. Specifically, in the proposed learning scheme, we design an adaptive utility function and an adaptive learning step for the aforementioned two problems, respectively. Empirical studies show that our algorithm achieves significantly better results than the state-of-the-art method on two real-world data sets.

Matrix factorization (MF) has been proved a very successful technique for Collaborative Filtering (CF), and hence has been widly adpoted in today's recommender systems. However, many studies have been proved that MF alone is poor to reveal the local relationships of users and items which can be learned well by the neighborhood-aware methods. To combine the merits of both approaches, in this paper, we propose a novel model which can effectively integrate the local preference information into MF. Different from various proposed methods which focus on representing the local similarity by the interactions of corresponding latent factors, we extend the neighborhood relationships to both latent factors and their rating preference. First, we establish clusters of users and items based on neighborhood information. Second, we transform the cluster information into two rating matrices which represent (user cluster) - (item) and (user) - (item cluster) preference. Third, we combine the generated rating matrices and the local latent factors into a single model, named Similarity-Based Matrix Factorization (SBMF). Since our model can explore the external representation of similarity information, it leads to more accurate recommendations. Experimental results on several real-world data sets show that our SBMF outperforms the state-of-the-art methods.

Recommendation as a universal service has driven much research works, among which explicit feedback estimation (e.g., Rating prediction in the Netflix competition) is probably the most well-known and well-studied problem. However, in various online and mobile applications, data resources of implicit feedbacks from users' interaction behaviors and linked connections from pervasive social media sites are more abundant. In this paper, we aim to integrate the users' implicit feedbacks and social connections in order to improve the ranking-oriented recommendation performance. One fundamental challenge is the noise of the social connections, which may cause incorrect social influences during learning of users' preferences. As a response, we propose to learn social experts (rather than to rely on connected individual users) as the major influence source for a certain user, which is likely to generate more accurate social influences. Specifically, we design a novel user preference generation function so as to seamlessly incorporate influences from the learned social experts. We then develop a general learning algorithm correspondingly, i.e., Collaborative ranking via learning social experts (CRSE). To verify our idea of learning social experts, we study the ranking performance of CRSE on two real-world datasets, and find that it can produce more accurate recommendations than the state-of-the-art methods.

As a state-of-the-art recommendation technique, collaborative filtering (CF) methods compute recommendations by leveraging a historical data set of users' ratings for items. So far, the best performing CF methods are latent factor models. Probabilistic matrix factorization (PMF) model, as a widely used latent factor model, offers a probabilistic foundation for regularization. In this paper, we present a novel CF method by incorporating implicit relationship between items into the basic PMF model. Firstly we mine the implicit correlation between items based on a matrix factorization model by utilizing contextual information, and then generalize recommendations by incorporating the obtained item relationship into the basic PMF model. We validate our approach on two datasets, and the experimental results show that the proposed method outperforms several existing CF models.

In the existing methods for solving matrix completion, such as singular value thresholding (SVT), soft-impute and fixed point continuation (FPCA) algorithms, it is typically required to repeatedly implement singular value decompositions (SVD) of matrices.When the size of the matrix in question is large, the computational complexity of finding a solution is costly. To reduce this expensive computational complexity, we apply Kronecker products to handle the matrix completion problem. In particular, we propose using Kronecker factorization, which approximates a matrix by the Kronecker product of several matrices of smaller sizes. Weintroduce Kronecker factorization into the soft-impute framework and devise an effective matrix completion algorithm.Especially when the factorized matrices have about the samesizes, the computational complexity of our algorithm is improved substantially.

As one of the most popular recommender system technologies, neighborhood-based collaborative filtering algorithm has obtained great favor due to its simplicity, justifiability, and stability. However, when faced with large-scale, sparse, or noise affected data, nearest-neighbor collaborative filtering performs not so well, as the calculation of similarity between user or item pairs is costly and the accuracy of similarity can be easily affected by noise and sparsity. In this paper, we present a novel collaborative filtering method based on user stars. Instead of treating every user as the same, we propose a method to generate a small number of users as the most reliable \emph{star users} and then produce predictions for the general population based on star users' ratings. Empirical studies on two different datasets suggest that our method outperforms traditional neighborhood-based collaborative filtering algorithm in terms of both efficiency and accuracy.

Classification can be regarded as dividing the data space into decision regions separated by decision boundaries. In this paper we analyze decision tree algorithms and the NBTree algorithm from this perspective. Thus, a decision tree can be regarded as a classifier tree, in which each classifier on a non-root node is trained in decision regions of the classifier on the parent node. Meanwhile, the NBTree algorithm, which generates a classifier tree with the C4.5 algorithm and the naive Bayes classifier as the root and leaf classifiers respectively, can also be regarded as training naive Bayes classifiers in decision regions of the C4.5 algorithm. We propose a second division (SD) algorithm and three soft second division (SD-soft) algorithms to train classifiers in decision regions of the naive Bayes classifier. These four novel algorithms all generate two-level classifier trees with the naive Bayes classifier as root classifiers. The SD and three SD-soft algorithms can make good use of both the information contained in instances near decision boundaries, and those that may be ignored by the naive Bayes classifier. Finally, we conduct experiments on 30 data sets from the UC Irvine (UCI) repository. Experiment results show that the SD algorithm can obtain better generalization abilities than the NBTree and the averaged one-dependence estimators (AODE) algorithms when using the C4.5 algorithm and support vector machine (SVM) as leaf classifiers. Further experiments indicate that our three SD-soft algorithms can achieve better generalization abilities than the SD algorithm when argument values are selected appropriately.

As compared with text spam, the image spam is a variant which is invented to escape from traditional text-based spam classification and filtering. Various approaches to image spam filtering have been proposed with respective advantages and drawbacks in terms of time cost and efficiency. In this paper, we propose a new approach based on Base64 encoding of image files and n-gram technique for feature extraction. By transforming normal images into Base64 presentation, we try to extract features of an image with n-gram technique. With these features we train an SVM (support vector machine) which shows effectiveness and efficiency in detecting spam images from legitimate images. With an online shared personal corpus of images as the input, experimental results show that our approach, in comparison with some of the existing methods of feature extraction, can achieve very high performance for image spam classification in terms of some basic measures such as accuracy, precision, and recall. Moreover, our approach shows its practicability by taking less running time for image spam classification in comparison to other methods.

Data perturbation is a popular method to achieve privacy-preserving data mining. However, distorted databases bring enormous overheads to mining algorithms as compared to original databases. In this paper, we present the GrC-FIM algorithm to address the efficiency problem in mining frequent itemsets from distorted databases. Two measures are introduced to overcome the weakness in existing work: firstly, the concept of independent granule is introduced, and granule inference is used to distinguish between non-independent itemsets and independent itemsets. We further prove that the support counts of non-independent itemsets can be directly derived from subitemsets, so that the error-prone reconstruction process can be avoided. This could improve the efficiency of the algorithm, and bring more accurate results; secondly, through the granular-bitmap representation, the support counts can be calculated in an efficient way. The empirical results on representative synthetic and real-world databases indicate that the proposed GrC-FIM algorithm outperforms the popular EMASK algorithm in both the efficiency and the support count reconstruction accuracy.

In order to preserve individual privacy, original data is distorted with the perturbation technique, and with the support reconstruction method, frequent itemsets can be mined from the distorted database. Due to this, mining process can be apart from being error-prone, expensively, in the dynamic update environment, more expensive in terms of time as compared to the original database. Some methods proposed try to solve this problem, but still not efficient. To improve so, this paper makes use of a method based on granular computing (GrC) in incremental mining, which is efficient and accuracy in support computation. The experiment results show the efficiency of our algorithm

The design of user friendly and expressive virtual brush systems for interactive digital painting and calligraphy has attracted a lot of attention and effort in both computer graphics and human-computer interaction circles for a long time. Providing a digital environment for paper-less artwork creation is not only challenging in terms of algorithmic design, but also promising for its potential market values. This paper proposes a novel algorithmic framework for interactive digital painting and calligraphy based a novel virtual hairy brush model. The algorithms in the kernel of our simulation framework are built upon solid modeling techniques. Implementing the algorithms, we have developed a virtual hairy brush prototype system with which end users can interactively produce high-quality digital paintings and calligraphic artwork. (The latest progress of our virtual brush project is reported at the website "http://www.cs.hku.hk/~songhua/e-brush/".) Copyright by Science in China Press 2005.

Better information leads to better design. In landscape design, the designers care for what plans affect the urban more than what their plans look like. In this paper, we discuss some key techniques in a GIS based landscape design system. With the advantage of the GIS, designers can make better decisions. A prototype named landscape evaluation system (LES) has been implemented. In LES system, designers can build up their virtual evaluation world, and establish the interrelationship of the entities in their virtual world.