Xudong Mao

Least Squares Generative Adversarial Networks

Unsupervised learning with generative adversarial networks (GANs) has proven hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. However, we found that this loss function may lead to the vanishing gradients problem during the learning process. To overcome such a problem, we propose in this paper the Least Squares Generative Adversarial Networks (LSGANs) which adopt the least squares loss function for the discriminator. We show that minimizing the objective function of LSGAN yields minimizing the Pearson X2 divergence. There are two benefits of LSGANs over regular GANs. First, LSGANs are able to generate higher quality images than regular GANs. Second, LSGANs perform more stable during the learning process. We evaluate LSGANs on LSUN and CIFAR-10 datasets and the experimental results show that the images generated by LSGANs are of better quality than the ones generated by regular GANs. We also conduct two comparison experiments between LSGANs and regular GANs to illustrate the stability of LSGANs.

2014 Special Issue: Community-aware user profile enrichment in folksonomy

In the era of big data, collaborative tagging (a.k.a. folksonomy) systems have proliferated as a consequence of the growth of Web 2.0 communities. Constructing user profiles from folksonomy systems is useful for many applications such as personalized search and recommender systems. The identification of latent user communities is one way to better understand and meet user needs. The behavior of users is highly influenced by the behavior of their neighbors or community members, and this can be utilized in constructing user profiles. However, conventional user profiling techniques often encounter data sparsity problems as data from a single user is insufficient to build a powerful profile. Hence, in this paper we propose a method of enriching user profiles based on latent user communities in folksonomy data. Specifically, the proposed approach contains four sub-processes: (i) tag-based user profiles are extracted from a folksonomy tripartite graph; (ii) a multi-faceted folksonomy graph is constructed by integrating tag and image affinity subgraphs with the folksonomy tripartite graph; (iii) random walk distance is used to unify various relationships and measure user similarities; (iv) a novel prototype-based clustering method based on user similarities is used to identify user communities, which are further used to enrich the extracted user profiles. To evaluate the proposed method, we conducted experiments using a public dataset, the results of which show that our approach outperforms previous ones in user profile enrichment.

Context-Aware personalized search based on user and resource profiles in folksonomies

The explosion of collaborative tagging data nowadays prompts an urgent demand upon Web 2.0 communities in assisting users to search interested resources quickly and effectively. Such a requirement entails much research on utilization of tag-based user and resource profiles so as to provide a personalized search in folksonomies. However, one major shortage for existing methods is their uniform treatment of user profile in the same way for each query, hence the search context for each query is ignored. In this paper, we focus on addressing this problem by modeling the search context. To capture and understand user intention, a nested context model is proposed. Furthermore, we conduct the experimental evaluation upon a real life data set, and the experimental result demonstrates that our approach is more effective than baselines.

Finding Dominating Set from Verbal Contextual Graph for Personalized Search in Folksonomy

With the development of the Internet, user-generated data has been growing tremendously in Web 2.0 era. Facing such a big volume of resources in folksonomy, people need a method of fast exploration and indexing to find their demanded data. To achieve this goal, contextual information is indispensable and valuable to understand user preference and purpose. In sociolinguistics, context can be mainly categorized as verbal context and social context. Comparing with verbal context, social context not only requires domain knowledge to pre-define contextual attributes but also acquires additional data from users. However, there is no research of addressing irrelevant contextual factors for verbal context model so far. The dominating set from verbal context proposed in this paper is to fill this blank. We present the verbal context in folksonomy to capture the user intention, and propose a dominating set discovering method for this verbal context model to prune the irrelevant contextual factors and keep the major characteristics at the same time. Furthermore, the experiments, which are conducted on a public data set, show that the proposed method gives convincing results.

Popularity Tendency Analysis of Ranking-Oriented Collaborative Filtering from the Perspective of Loss Function

Collaborative filtering (CF) has been the most popular approach for recommender systems in recent years. In order to analyze the property of a ranking-oriented CF algorithm directly and be able to improve its performance, this paper investigates the ranking-oriented CF from the perspective of loss function. To gain the insight into the popular bias problem, we also study the tendency of a CF algorithm in recommending the most popular items, and show that such popularity tendency can be adjusted through setting different parameters in our models. After analyzing two state-of-the-art algorithms, we propose in this paper two models using the generalized logistic loss function and the hinge loss function, respectively. The experimental results show that the proposed methods outperform the state-of-the-art algorithms on two real data sets.

Recent advances in semantic computing and personalization

h 0 With the rapid growth of social communities and intelligent eb services in recent years, there has been a huge volume of ser-generated data in the internet every day [1] . To exploit such large collection of web data, it is essential to identify the unerlying high-level semantics of data in multiple modalities and ources. Such semantically-rich information facilitates the undertanding of user intentions, needs and preferences. Furthermore, he paradigm of personalized information access, which has been ncreasingly employed in various mobile and web-based systems o avoid information overload and better satisfy end users’ inforation needs, is supported by various semantic computing techiques such as ontology [2] , user profiling [3] , social annotations 4] and so on. Many research questions and challenges need to be ddressed for the realization of powerful personalized models with emantic computing techniques. This special issue aims to investigate (i) how high-level semanics are extracted and exploited from web and social data sources ia state-of-the-art data mining techniques (e.g., support vector achines [5] , Adaboost [6] , deep neural networks [7] , etc) and (ii) ow personalized models are facilitated and supported by underlyng semantics. In addition, the research issues in applying semantic omputing and personalization in web-based social communities nd interactive platforms are covered and discussed. To balance the quality and coverage of user reviews, ‘More Fous on What You Care About: Personalized Top Reviews Set’ resents a review recommendation model which identifies imporant aspects of the review and selects a top personalized reviews et according to user preferences. The effectiveness of the proosed model is further verified by conducting experiments on the atasets crawled from two e-commerce sites (i.e., Yelp and TripAdisor). In ‘Real-time Personalized Twitter Search based on Semanic Expansion and Quality Model’ , authors propose a personalized earch framework for real-time twitter data stream by implementng the semantic expansion based on user preferences and employng the quality model based on social features. The evaluation is ased on a real twitter data stream consisted of 51,770,318 tweets. he experimental findings showed that the framework can improve anking effectiveness and identify user preferences appropriately. By analyzing the semantic features in multi-aspect vocal ratings nd karaoke machine ratings, a personalized song recommender ystem based on a joint model is proposed in the article ‘Karaoke ong Recommendation Using Multiple Kernel Learning Approxmation . Specifically, the latent features of the vocal ratings are earnt by a multiple kernel learning method and then fed into the

Interactive tutoring of cooking activities with personalized multimedia recipe search

With the fast development of the Internet, there are more and more learning resources including various multimedia resources on the Web. Meanwhile, e-learning becomes more popular and important due to its convenience and autonomy. However, how to help users conduct high quality resource retrieval remains a challenge and should be considered first for an e-learning system. The traditional indexing approach which classifies multimedia resources into predefined categories can hardly meet user demands. Collaborative tagging (also known as folksonomy) systems provide users with a simple but powerful mechanism to obtain required multimedia resources. Moreover, feedbacks made by users can be utilized to refine the retrieval results during the interactive process between users and systems. In this paper, we present an interactive tutoring system with personalized multimedia resource search, based on collaborative tagging mechanism. We also describe several application scenarios in which content-based image retrieval (CBIR) is combined with collaborative tagging to facilitate interactive tutoring in our cooking recipe database system.