scholarly journals Attention-Enhanced Graph Neural Networks for Session-Based Recommendation

Mathematics ◽  
2020 ◽  
Vol 8 (9) ◽  
pp. 1607 ◽  
Author(s):  
Baocheng Wang ◽  
Wentao Cai

Session-based recommendation, which aims to match user needs with rich resources based on anonymous sessions, nowadays plays a critical role in various online platforms (e.g., media streaming sites, search and e-commerce). Existing recommendation algorithms usually model a session as a sequence or a session graph to model transitions between items. Despite their effectiveness, we would argue that the performance of these methods is still flawed: (1) Using only fixed session item embedding without considering the diversity of users’ interests and target items. (2) For user’s long-term interest, the difficulty of capturing the different priorities for different items accurately. To tackle these defects, we propose a novel model which leverages both the target attentive network and self-attention network to improve the graph-neural-network (GNN)-based recommender. In our model, we first model user’s interaction sequences as session graphs which serves as the input of the GNN, and each node vector involved in session graph can be obtained via the GNN. Next, target attentive network can activates different user interests corresponding to varied target items (i.e., the session embedding learned varies with different target items), which can reveal the relevance between users’ interests and target items. At last, after applying the self-attention mechanism, the different priorities for different items can be captured to improve the precision of the long-term session representation. By using a hybrid of long-term and short-term session representation, we can capture users’ comprehensive interests at multiple levels. Extensive experiments demonstrate the effectiveness of our algorithm on two real-world datasets for session-based recommendation.

2020 ◽  
Vol 34 (04) ◽  
pp. 5045-5052 ◽  
Author(s):  
Chen Ma ◽  
Liheng Ma ◽  
Yingxue Zhang ◽  
Jianing Sun ◽  
Xue Liu ◽  
...  

The chronological order of user-item interactions can reveal time-evolving and sequential user behaviors in many recommender systems. The items that users will interact with may depend on the items accessed in the past. However, the substantial increase of users and items makes sequential recommender systems still face non-trivial challenges: (1) the hardness of modeling the short-term user interests; (2) the difficulty of capturing the long-term user interests; (3) the effective modeling of item co-occurrence patterns. To tackle these challenges, we propose a memory augmented graph neural network (MA-GNN) to capture both the long- and short-term user interests. Specifically, we apply a graph neural network to model the item contextual information within a short-term period and utilize a shared memory network to capture the long-range dependencies between items. In addition to the modeling of user interests, we employ a bilinear function to capture the co-occurrence patterns of related items. We extensively evaluate our model on five real-world datasets, comparing with several state-of-the-art methods and using a variety of performance metrics. The experimental results demonstrate the effectiveness of our model for the task of Top-K sequential recommendation.


2020 ◽  
Vol 34 (01) ◽  
pp. 214-221 ◽  
Author(s):  
Ke Sun ◽  
Tieyun Qian ◽  
Tong Chen ◽  
Yile Liang ◽  
Quoc Viet Hung Nguyen ◽  
...  

Point-of-Interest (POI) recommendation has been a trending research topic as it generates personalized suggestions on facilities for users from a large number of candidate venues. Since users' check-in records can be viewed as a long sequence, methods based on recurrent neural networks (RNNs) have recently shown promising applicability for this task. However, existing RNN-based methods either neglect users' long-term preferences or overlook the geographical relations among recently visited POIs when modeling users' short-term preferences, thus making the recommendation results unreliable. To address the above limitations, we propose a novel method named Long- and Short-Term Preference Modeling (LSTPM) for next-POI recommendation. In particular, the proposed model consists of a nonlocal network for long-term preference modeling and a geo-dilated RNN for short-term preference learning. Extensive experiments on two real-world datasets demonstrate that our model yields significant improvements over the state-of-the-art methods.


Brain-Mind ◽  
2019 ◽  
pp. 252-276
Author(s):  
Paul Thagard

The self is a complex of mechanisms at multiple levels that include the molecular and the social. Semantic pointers are crucial to the self with respect to various phenomena, including how one represents oneself to oneself and to others, as well as in how one evaluates oneself. Also explained are operations that the self does to itself in efforts to achieve short-term goals such as self-control and long-term goals such as self-fulfillment. Semantic pointer explanations of images, concepts, and other mental representations are important for understanding how selves accomplish their goals. Representations of the self via semantic pointers can recursively be bound into semantic pointers for beliefs, desires, and intentions. Discussion of the social mechanisms relevant to the self begins to connect neural and mental mechanisms with discussions of social sciences and professions.


2021 ◽  
pp. 1-5
Author(s):  
Kalliopi Megari ◽  
Kalliopi Megari

Background and Objective: Postoperative cognitive dysfunction (POCD) involves decline in several cognitive domains after surgery and is particularly common after cardiac surgery. Given the potential effects of such cognitive dysfunction on quality of life, it is important to study it in multiple populations in order to limit its occurrence. Recent advances in surgical technology may assist in achieving this goal. Methods: We present the long-term neuropsychological outcome of two elderly patients, one of whom had off pump heart surgery and the other oncological surgery. We administered a series of neuropsychological tests assessing attention, complex scanning, verbal working memory, executive functioning, short-term and long-term memory, and visuospatial perception before surgery, prior to discharge, at 3-month follow-up and 6 years after surgery. We compared the performance of these two patients to normative datasets. Results: Despite equivalent levels of pre-surgery performance between the two patients, the oncology patient exceeded his preoperative neurocognitive levels, suggesting less postoperative cognitive dysfunction in the heart patient overall, on all neuropsychological domains at 6-year follow-up, except short-term retention. In contrast, the heart patient showed no improvement, and, instead, showed some cognitive decline which remained consistent over time. Conclusion: Our findings highlight the critical role of the type of surgery utilized in the development of POCD and have implications for clinical management and patients’ quality of life in the very long term.


Blood ◽  
2012 ◽  
Vol 120 (21) ◽  
pp. 1112-1112
Author(s):  
Cornelia Fischer ◽  
Brigitte Spath ◽  
Ali Amirkhosravi ◽  
Walter Fiedler ◽  
Carsten Bokemeyer ◽  
...  

Abstract Abstract 1112 Acute myelogenous leukemia (AML) may be complicated by DIC. TF plays a critical role in AML-associated coagulopathy, and induction of apoptosis significantly increases TF PCA on leukemic blasts, mainly via phosphatidylserine (PS) membrane exposure. However, PDI, a thiol isomerase with oxidoreductase and chaperone activity, has also been implicated in cellular TF regulation. Particularly, PDI inhibitors have been shown to exert antithrombotic activity in animal models. Besides its predominant localization in the endoplasmic reticulum, PDI is present on cell surfaces, where it may represent a promising therapeutic target. We investigated the effect of PDI inhibitors on the expression of TF PCA by leukemic HL60 and THP1 cells to explore their potential as anticoagulant drugs for the prevention and/or treatment of AML-associated DIC. Using a fluorescence-based insulin reduction assay, we confirmed inhibition of recombinant human PDI by bacitracin and quercetin-3-rutinoside (also known as rutin and recently shown to be a specific PDI inhibitor) with IC50 values of 0.6 mM and 14 μM, respectively, showing >95% inhibition at 1 mM (bacitracin) and 50 μM (rutin). Significant insulin reductase activity was observed on HL60 cells, and this activity was inhibited by 75% and 49% using 1 mM bacitracin and 100 μM rutin, respectively, suggesting the presence of additional, PDI-independent thiol isomerase activity. Short-term treatment with 100 μM rutin for 15 min also inhibited TF PCA on HL60 cells by 37%. Importantly, the inhibitory effect of rutin on cell-associated PDI and TF activity was completely abolished by cell washing, confirming previous evidence that rutin is a reversible PDI inhibitor. When HL60 cells were exposed to rutin (100 μM) for 24 hrs, cell-associated TF PCA was increased 2.3-fold (P<0.01), an effect that was accompanied by enhanced PS exposure, as assessed by annexin V-FITC binding (positive cells, 32±11 vs. 10±4%; P<0.01), and increased PCA of cellular microparticles (MPs) isolated from culture supernatants, as evidenced by the thrombin generation parameters lag phase (LP, 14±1 vs. 19±4 min), peak thrombin (PT, 55±17 vs. 22±14 nM), and area under the curve (AUC, 1193±329 vs. 476±347 nM*min; P<0.01). Interestingly, treatment with 100 μM rutin also resulted in a 1.7-fold increase in total cellular TF antigen (P=0.07). The effects of long-term incubation with bacitracin (1 mM) were even more pronounced, involving an 8.3-fold and 4.6-fold increase in cell-associated TF PCA and total cellular TF antigen, respectively. PS exposure (45±9%) and shedding of procoagulant MPs (LP, 7±1 min; PT, 175±49 nM; AUC, 2756±402 nM*min) were also significantly increased. While neither short-term nor long-term exposure to rutin affected TF PCA on THP1 cells, co-incubation with rutin dose-dependently (10–100 μM) inhibited daunorubicin-induced TF PCA in this cell model, an effect that could not be explained by decreased PS exposure. Importantly, both the reaction pattern of HL60 and that of THP1 cells were reproduced ex vivo using myeloblasts from AML patients. In summary, our findings suggest a highly complex and context-dependent role of PDI in leukemic-cell TF PCA expression. While short-term exposure to rutin can reversibly inhibit both PDI and TF activity, long-term exposure may result in significantly increased cellular TF PCA and MP shedding, pointing to a possible role of PDI in PS homeostasis, cytoskeleton rearrangement, and/or TF recycling. In addition, induction of leukemic-cell apoptosis and necrosis by cytotoxic drugs, which is associated with an early loss in membrane integrity and enhanced accessibility of cytoplasmic enzymes, may involve an additional role of (intracellular) PDI in the efficient presentation of TF PCA by AML blasts. Disclosures: No relevant conflicts of interest to declare.


Sensors ◽  
2020 ◽  
Vol 20 (10) ◽  
pp. 2832
Author(s):  
Nazanin Fouladgar ◽  
Kary Främling

Multivariate time series with missing data is ubiquitous when the streaming data is collected by sensors or any other recording instruments. For instance, the outdoor sensors gathering different meteorological variables may encounter low material sensitivity to specific situations, leading to incomplete information gathering. This is problematic in time series prediction with massive missingness and different missing rate of variables. Contribution addressing this problem on the regression task of meteorological datasets by employing Long Short-Term Memory (LSTM), capable of controlling the information flow with its memory unit, is still missing. In this paper, we propose a novel model called forward and backward variable-sensitive LSTM (FBVS-LSTM) consisting of two decay mechanisms and some informative data. The model inputs are mainly the missing indicator, time intervals of missingness in both forward and backward direction and missing rate of each variable. We employ this information to address the so-called missing not at random (MNAR) mechanism. Separately learning the features of each parameter, the model becomes adapted to deal with massive missingness. We conduct our experiment on three real-world datasets for the air pollution forecasting. The results demonstrate that our model performed well along with other LSTM-derivation models in terms of prediction accuracy.


2013 ◽  
Vol 427-429 ◽  
pp. 2368-2372
Author(s):  
Wen Juan Lian ◽  
Jian Li Zhao ◽  
Yu Jun Li

The paper introduces a personalized intelligent recommender system. The system characters include: (1) integrating Agent technology to improve the recommender system reactivity, proactivity and autonomy; (2) giving a mixed recommendation method with content-based recommendation; (3) proposing the personalized user model, which can describe long-term interests and short-term interests, effectively deal with the user interests drift problem.


2015 ◽  
Vol 26 (09) ◽  
pp. 1550102 ◽  
Author(s):  
Wen-Jun Li ◽  
Yuan-Yuan Xu ◽  
Qiang Dong ◽  
Jun-Lin Zhou ◽  
Yan Fu

Traditional recommender algorithms usually employ the early and recent records indiscriminately, which overlooks the change of user interests over time. In this paper, we show that the interests of a user remain stable in a short-term interval and drift during a long-term period. Based on this observation, we propose a time-aware diffusion-based (TaDb) recommender algorithm, which assigns different temporal weights to the leading links existing before the target user's collection and the following links appearing after that in the diffusion process. Experiments on four real datasets, Netflix, MovieLens, FriendFeed and Delicious show that TaDb algorithm significantly improves the prediction accuracy compared with the algorithms not considering temporal effects.


Physiology ◽  
2009 ◽  
Vol 24 (1) ◽  
pp. 17-25 ◽  
Author(s):  
Theresa Berndt ◽  
Rajiv Kumar

Phosphorus plays a critical role in diverse biological processes, and, therefore, the regulation of phosphorus balance and homeostasis are critical to the well being of the organism. Changes in environmental, dietary, and serum concentrations of inorganic phosphorus are detected by sensors that elicit changes in cellular function and alter the efficiency by which phosphorus is conserved. Short-term, post-cibal responses that occur independently of hormones previously thought to be important in phosphorus homeostasis may play a larger role than previously appreciated in the regulation of phosphorus homeostasis. Several hormones and regulatory factors such as the vitamin D endocrine system, parathyroid hormone, and the phosphatonins (FGF-23, sFRP-4, MEPE) among others, may play a role only in the long-term regulation of phosphorus homeostasis. In this review, we discuss how organisms sense changes in phosphate concentrations and how changes in hormonal factors result in the conservation or excretion of phosphorus.


2021 ◽  
Author(s):  
Haomei Duan ◽  
Jinghua Zhu

In the case that user profiles are not available, the recommendation based on anonymous session is particularly important, which aims to predict the items that the user may click at the next moment based on the user's access sequence over a while. In recent years, with the development of recurrent neural network, attention mechanism, and graph neural network, the performance of session-based recommendation has been greatly improved. However, the previous methods did not comprehensively consider the context dependencies and short-term interest first of the session. Therefore, we propose a context-aware short-term interest first model (CASIF).The aim of this paper is improve the accuracy of recommendations by combining context and short-term interest. In CASIF, we dynamically construct a graph structure for session sequences and capture rich context dependencies via graph neural network (GNN), latent feature vectors are captured as inputs of the next step. Then we build the shortterm interest first module, which can to capture the user's general interest from the session in the context of long-term memory, at the same time get the user's current interest from the item of the last click. In the end, the short-term and long-term interest are combined as the final interest and multiplied by the candidate vector to obtain the recommendation probability. Finally, a large number of experiments on two real-world datasets demonstrate the effectiveness of our proposed method.


Sign in / Sign up

Export Citation Format

Share Document