The huge volume of emerging graph datasets has become a double-bladed sword for machine learning. On the one hand, it empowers success myriad neural networks (GNNs) with strong empirical performance. other training modern on data is computationally expensive. How to distill given dataset while retaining most trained models' performance challenging problem. Existing efforts try approach this problem by solving meta-learning-based bilevel optimization objectives. A major hurdle lies in that...

Knowledge graph question answering aims to identify answers of the query according facts in knowledge graph. In vast majority existing works, input queries are considered perfect and can precisely express user's intention. However, reality, might be ambiguous elusive which only contain a limited amount information. Directly these may yield unwanted deteriorate user experience. this paper, we propose PReFNet focuses on with pseudo relevance feedback graphs. order leverage hidden (pseudo)...

Contrastive Learning (CL) has shown promising performance in collaborative filtering. The key idea is to generate augmentation-invariant embeddings by maximizing the Mutual Information between different augmented views of same instance. However, we empirically observe that existing CL models suffer from \textsl{dimensional collapse} issue, where user/item only span a low-dimension subspace entire feature space. This suppresses other dimensional information and weakens distinguishability...

Network embedding plays an important role in a variety of social network applications. Existing methods, explicitly or implicitly, can be categorized into positional (PE) methods structural (SE) methods. Specifically, PE encode the information and obtain similar embeddings for adjacent/close nodes, while SE aim to learn identical representations nodes with same local patterns, even if two are far away from each other. The disparate designs types lead apparent dilemma that no could perfectly...

Collaborative rating sites have become essential resources that many users consult to make purchasing decisions on various items. Ideally, a user wants quickly decide whether an item is desirable, especially when choices are available. In practice, however, either spends lot of time examining reviews before making informed decision, or simply trusts overall aggregations associated with item. this paper, we argue neither option satisfactory and propose novel powerful third option, Meaningful...

Knowledge discovery from temporal, spatial and spatio-temporal data is pivotal for understanding predicting the behavior of Earth's ecosystem model. An important influence leaving its impact on global climate system. In this paper, Earth Science that we have analyzed consists daily air temperature precipitation measurements, aggregated heterogeneous sensors fifty years (1950--1999). The enormous amount available analysis requires employment mining techniques discovering interesting patterns,...

Graph Neural Networks (GNNs) combine node attributes over a fixed granularity of the local graph structure around to predict its label. However, different nodes may relate node-level property with neighborhood, and using same level smoothing for all can be detrimental their classification. In this work, we challenge common fact that single GNN layer classify by training GNNs distinct personalized each node. Inspired metric learning, propose novel algorithm, MetSelect1, select optimal...

Recent years have witnessed a surge of interest in machine learning on graphs and networks with applications ranging from IoT traffic management to social network recommendations. Supervised tasks such as node classification link prediction require us perform feature engineering that is known agreed be the key success applied learning. Research efforts dedicated representation learning, especially using deep has shown ways automatically learn relevant features vast amounts potentially noisy,...

The rapid proliferation of new users and items on the social web has aggravated gray-sheep user/long-tail item challenge in recommender systems. Historically, cross-domain co-clustering methods have successfully leveraged shared across dense sparse domains to improve inference quality. However, they rely rating data cannot scale multiple target (i.e., one-to-many transfer setting). This, combined with increasing adoption neural architectures, motivates us develop scalable layer-transfer...

The rise of Web 2.0 is signaled by sites such as Flickr, del.icio.us, and YouTube, social tagging essential to their success. A typical action involves three components, user, item (e.g., photos in Flickr), tags (i.e., words or phrases). Analyzing how are assigned certain users items has important implications helping search for desired information. In this paper, we explore common analysis tasks propose a dual mining framework behavior mining. This centered around two opposing measures,...

We present ShoppingAdvisor, a novel recommender system that helps users in shopping for technical products. ShoppingAdvisor leverages both user preferences and product attributes order to generate its suggestions. The elicits via tree-shaped flowchart, where each node is question the user. At node, suggests ranking of products matching user, gets progressively refined along path from tree's root one leafs.

Collaborative rating sites such as IMDB and Yelp have become rich resources that users consult to form judgments about choose from among competing items. Most of these either provide a plethora information for interpret all by themselves or simple overall aggregate information. Such aggregates (e.g., average over who rated an item, along pre-defined dimensions, etc.) can not help user quickly decide the desirability item. In this paper, we build system MapRat allows explore multiple...

Collaborative Filtering (CF) has been successfully used to help users discover the items of interest. Nevertheless, existing CF methods suffer from noisy data issue, which negatively impacts quality recommendation. To tackle this problem, many prior studies leverage adversarial learning regularize representations users/items, improves both generalizability and robustness. Those often learn perturbations model parameters under min-max optimization framework. However, there still have two...

The popularity of collaborative tagging sites has created new challenges and opportunities for designers web items, such as electronics products, travel itineraries, popular blogs, etc. An increasing number people are turning to online reviews user-specified tags choose from among competing items. This creates an opportunity build items that likely attract desirable when published. In this paper, we consider a novel optimization problem: given training dataset existing with their...

Graph Neural Networks (GNNs) have achieved impressive performance in collaborative filtering. However, recent studies show that GNNs tend to yield inferior when the distributions of training and test data are not aligned well. Moreover, often requires optimizing non-convex neural networks with an abundance local global minima, which may differ widely their at time. Thus, it is essential develop optimization strategy can choose minima carefully, strong generalization on unseen data. Here we...

The increasing popularity and widespread use of online review sites over the past decade has motivated businesses all types to possess an expansive arsenal user feedback (preferably positive) in order mark their reputation presence Web. Though a significant proportion purchasing decisions today are driven by average numeric scores (e.g., movie rating IMDB), detailed reviews critical for activities such as buying expensive digital SLR camera, reserving vacation package, etc. Since writing...

The scale and significance of graph structured data today has led to the development management systems that are optimized either for navigation requests or analytic requests. We present a general purpose system provides high performance concurrently both In addition, it supports highly dynamic graphs wherein vertices edges added deleted properties modified. Our solution employs hybrid architecture comprising two engines, one each workload, with synchronization unit manage updates federation...

The finance industry is constantly faced with an ever evolving set of challenges including credit card fraud, identity theft, network intrusion, money laundering, human trafficking, and illegal sales firearms. There also the newly emerging threat fake news in financial media that can lead to distortions trading strategies investment decisions. In addition, traditional problems such as customer analytics, forecasting, recommendations take on a unique flavor when applied data. A number new...

The rise of Web 2.0 is signaled by sites such as Flickr, del.icio.us, and YouTube, social tagging essential to their success. A typical action involves three components, user, item (e.g., photos in Flickr), tags (i.e., words or phrases). Analyzing how are assigned certain users items has important implications helping search for desired information. In this paper, we explore common analysis tasks propose a dual mining framework behavior mining. This centered around two opposing measures,...

Association rule mining is an indispensable tool for discovering insights from large databases and data warehouses.The in a warehouse being multi-dimensional, it often useful to mine rules over subsets of defined by selections the dimensions.Such interactive multi-dimensional query windows difficult since computationally expensive.Current methods using pre-computation frequent itemsets require counting some revisiting transaction database at time, which very expensive.We develop method (RMW)...

Recent years have witnessed a surge of interest in machine learning on graphs and networks with applications ranging from vehicular network design to IoT traffic management social recommendations. Supervised tasks such as node classification link prediction require us perform feature engineering that is known agreed be the key success applied learning. Research efforts dedicated representation learning, especially using deep has shown ways automatically learn relevant features vast amounts...

The widespread use and growing popularity of online collaborative content sites (e.g., Yelp, Amazon, IMDB) has created rich resources for users to consult in order make purchasing decisions on various items such as restaurants, e-commerce products, movies, etc. It also new opportunities producers improve business by designing better composing succinct advertisement snippets building smart personalized recommendation systems. This motivates us develop a framework exploratory mining user...