With the rapid prevalence of smart mobile devices and dramatic proliferation location-based social networks (LBSNs), recommendation has become an important means to help people discover attractive interesting points interest (POIs). However, extreme sparsity user-POI matrix cold-start issue create severe challenges, causing CF-based methods degrade significantly in their performance. Moreover, requires spatiotemporal context awareness dynamic tracking user's latest preferences a real-time manner.

End-to-end neural models for intelligent dialogue systems suffer from the problem of generating uninformative responses. Various methods were proposed to generate more informative responses by leveraging external knowledge. However, few previous work has focused on selecting appropriate knowledge in learning process. The inappropriate selection could prohibit model make full use Motivated this, we propose an end-to-end which employs a novel mechanism where both prior and posterior...

With the rapid development of location-based social networks (LBSNs), spatial item recommendation has become an important way helping users discover interesting locations to increase their engagement with services. Although human movement exhibits sequential patterns in LBSNs, most current studies on recommendations do not consider influence locations. Leveraging is, however, very challenging, considering 1) users' check-in data LBSNs a low sampling rate both space and time, which renders...

Classical recommender systems provide users with a list of recommendations where each recommendation consists single item, e.g., book or DVD. However, several applications can benefit from system capable recommending packages items, in the form sets. Sample include travel planning limited budget (price time) and twitter wanting to select worthwhile tweeters follow given that they deal only bounded number tweets. In these contexts, there is need for recommend top-k user choose from.

Extracting interesting tuples from a large database is an important problem in multi-criteria decision making. Two representative queries were proposed the literature: top- k and skyline queries. A query requires users to specify their utility functions beforehand then returns users. does not require any function but it puts no control on number of returned Recently, k-regret was received attention community because output size controllable, thus avoids those deficiencies Specifically, that...

Collaborative Filtering (CF) is the most popular method for recommender systems. The principal idea of CF that users might be interested in items are favorited by similar users, and existing methods measure users' preferences their behaviours over all items. However, have different interests topics, thus share with groups sets In this paper, we propose a novel scalable CCCF which improves performance via user-item co-clustering. first clusters into several subgroups, where each subgroup...

Existing recommender systems mostly focus on recommending individual items which users may be interested in. User-generated item lists the other hand have become a popular feature in many applications. E.g., Goodreads provides with an interface for creating and sharing interesting book lists. These user-generated complement main functionality of corresponding application, intuitively alternative way to browse discover consumed. Unfortunately, existing are not designed In this work, we study...

Abstract Fifteen adults with chronic low back pain (M - 4 years), age 18 to 43 years = 29 participated. All but one were moderately highly hypnotizable 7.87; modified 11-point Stanford Hypnotic Susceptibility Scale, Form C [Weitzenhoffer & Hilgard, 1962]), and significantly reduced perception following hypnotic analgesia instructions during cold-pressor training. In Part 1, somatosensory event-related potential correlates of noxious electrical stimulation evaluated attend (HA) conditions at...

Risk analysis and prioritization are a key process in project risk management (PRM). Its outcomes serve as input of the response planning where decisions made. Complexity projects is characterized by emergence phenomena that difficult to detect manage using classical methods. It may disturb assessment, on which priorities further established. This paper aims at importance measure (IM) techniques complex PRM field. involves modeling network providing complementary results based IMs accounting...

When faced with a database containing millions of tuples, an end user might be only interested in finding his/her (close to) favorite tuple the database. Recently, regret minimization query was proposed to obtain small subset from that fits user's needs, which are expressed through unknown utility function. Specifically, it minimizes "regret'' level user, we quantify as ratio if s/he gets best selected but not among all tuples We study how enhance interactions : when presented number (which...

Classical recommender systems provide users with a list of recommendations where each recommendation consists single item, e.g., book or DVD. However, applications such as travel planning can benefit from system capable recommending packages items, under user-specified budget and in the form sets sequences. In this context, there is need for that recommend top-k user to choose from. paper, we propose novel system, CompRec-Trip, which automatically generate composite planning. The leverages...

In product search, the retrieval of candidate products before re-ranking is more mission critical and challenging than other search like web especially for tail queries, which have a complex specific intent. this paper, we present hybrid system e-commerce deployed at Walmart that combines traditional inverted index embedding-based neural to better answer user queries. Our significantly improved relevance engine, measured by both offline online evaluations. The improvements were achieved...

There are several applications, such as play lists of songs or movies, and shopping carts, where users interested in finding top- k packages, consisting sets items. In response to this need, there has been a recent flurry activity around extending classical recommender systems (RS), which effective at recommending individual items, recommend The few proposals for package RS suffer from one the following drawbacks: they either rely on hard constraints may be difficult specified exactly by...

Large neural networks have aroused impressive progress in various real world applications. However, the expensive storage and computational resources requirement for running deep make them problematic to be deployed on mobile devices. Recently, matrix tensor decompositions been employed compressing networks. In this paper, we develop a simultaneous decomposition technique network optimization. The shared structure is first discussed. Sometimes, not only but also parameters are form...

Top-k query processing has recently received a significant amount of attention due to its wide application in information retrieval, multimedia search and recommendation generation. In this work, we consider the problem how efficiently answer top-k by using previously cached results. While there been some previous work on problem, existing algorithms suffer from either limited scope or lack scalability. paper, propose two novel for handling problem. The first algorithm LPTA+ provides...

Due to the prevalence of graph data, analysis is very important nowadays. One popular on data Random Walk with Restart (RWR) since it provides a good metric for measuring proximity two nodes in graph. Although RWR important, challenging design an algorithm RWR. To best our knowledge, there are no existing algorithms which, at same time, (1) index-free, (2) return answers theoretical guarantee and (3) efficient. Motivated by this, this paper, we propose index-free called Residue-Accumulated...

When a large dataset is given, it not desirable for user to read all tuples one-by-one in the whole find satisfied tuples. The traditional top-k query finds best k (i.e., tuples) w.r.t. user's preference. However, practice, difficult specify his/her preference explicitly. We study how enhance with interaction. Specifically, we ask several questions, each of which consists two and asks indicate one s/he prefers. Based on feedback, learned implicitly returned. Here, instead directly following...

When faced with a database containing millions of products, user may be only interested in (typically much) smaller representative subset. Various approaches were proposed to create good subset that fits the user's needs which are expressed form utility function (e.g., top-k and diversification query). Recently, regret minimization query was proposed: it does not require users provide their functions returns small set tuples such any favorite tuple this is guaranteed much worse than his/her...

For an E-commerce website like Walmart.com, search is one of the most critical channel for engaging customer. Most existing works on are composed two steps, a retrieval step which obtains candidate set matching items, and re-rank focuses fine-tuning ranking items. Inspired by latest in domain neural information (NIR), we discuss this work our exploration various product models trained log data. We lessons learned empirical result section, these results can be applied to any engine aims at...

End-to-end neural models for intelligent dialogue systems suffer from the problem of generating uninformative responses. Various methods were proposed to generate more informative responses by leveraging external knowledge. However, few previous work has focused on selecting appropriate knowledge in learning process. The inappropriate selection could prohibit model make full use Motivated this, we propose an end-to-end which employs a novel mechanism where both prior and posterior...