HyperSoRec: Exploiting Hyperbolic User and Item Representations with Multiple Aspects for Social-aware Recommendation

Social recommendation has achieved great success in many domains including e-commerce and location-based social networks. Existing methods usually explore the user-item interactions or user-user connections to predict users’ preference behaviors. However, they usually learn both user and item representations in Euclidean space, which has large limitations for exploring the latent hierarchical property in the data. In this article, we study a novel problem of hyperbolic social recommendation, where we aim to learn the compact but strong representations for both users and items. Meanwhile, this work also addresses two critical domain-issues, which are under-explored. First, users often make trade-offs with multiple underlying aspect factors to make decisions during their interactions with items. Second, users generally build connections with others in terms of different aspects, which produces different influences with aspects in social network. To this end, we propose a novel graph neural network (GNN) framework with multiple aspect learning, namely, HyperSoRec. Specifically, we first embed all users, items, and aspects into hyperbolic space with superior representations to ensure their hierarchical properties. Then, we adapt a GNN with novel multi-aspect message-passing-receiving mechanism to capture different influences among users. Next, to characterize the multi-aspect interactions of users on items, we propose an adaptive hyperbolic metric learning method by introducing learnable interactive relations among different aspects. Finally, we utilize the hyperbolic translational distance to measure the plausibility in each user-item pair for recommendation. Experimental results on two public datasets clearly demonstrate that our HyperSoRec not only achieves significant improvement for recommendation performance but also shows better representation ability in hyperbolic space with strong robustness and reliability.

Exploration of the Specificity of Motor Skills Hypothesis in 7–8 Year Old Primary School Children: Exploring the Relationship Between 12 Different Motor Skills From Two Different Motor Competence Test Batteries

This study examined the specificity hypothesis by examining the association between two specific motor competence test batteries [Movement Assessment Battery for Children (MABC) and Test of Motor Competence (TMC)] in a sample of young children. In addition, we explored the factorial structure of the MABC and TMC. A total of 80 children participated in the study (38 girls and 42 boys) with a mean chronological age of 7.9 years (SD 0.55). The correlation between total score MABC and total z-score TMC was r = 0.46. In general, low pair-wise correlations (r2 < 0.20) between the different motor tasks were found. The highest correlation was between the placing bricks and building bricks r = 0.45 (TMC); the stork balance and jumping in squares r = 0.45 (MABC). These low pair-wise relations of items are consistent with findings from younger and older children's age-related motor competence test batteries. Principal components analysis (PCA) showed that the 1st component accommodated 25% of the variance and was dominated in the top five variable weightings by items of the MABC test; whereas the 2nd component accommodated 12% of the variance with the higher weightings all from the TMC test. The findings provide evidence with children for specificity rather than generality in learning motor skills a viewpoint that has predominantly been driven by adult learning studies. The PCA revealed that the MABC and TMC are testing different properties of children's motor competence though in both cases the variance accounted for is relatively modest, but generally higher than the motor item pair-wise correlation.

Real-world size is automatically encoded in preschoolers’ object representations

When adults see a picture of an object, they automatically process how big the object typically is in the real world (Konkle & Oliva, 2012a). How much life experience is needed for this automatic size processing to emerge? Here, we ask whether preschoolers show this same signature of automatic size processing. We showed 3- and 4-year-olds displays with two pictures of objects and asked them to touch the picture that was smaller on the screen. Critically, the relative visual sizes of the objects could either be congruent with their relative real-world sizes (e.g., a small picture of a shoe next to a big picture of a car) or incongruent with their relative real-world sizes (e.g., a big picture of a shoe next to a small picture of a car). Across two experiments, we found that preschoolers were worse at making visual size judgments on incongruent trials, suggesting that real-world size was automatically activated and interfered with their performance. In a third experiment, we found that both 4-year-olds and adults showed similar item-pair effects (i.e., showed larger Size-Stroop effects for the pairs of items, relative to other pairs). Furthermore, the magnitude of the item-pair Stroop effects in 4-year-olds did not depend on whether they could recognize the pictured objects, suggesting that the perceptual features of these objects were sufficient to trigger the processing of real-world size information. These results indicate that, by 3–4 years of age, children automatically extract real-world size information from depicted objects.

Development and Initial Validation of an Inconsistent Responding Scale for the Youth Psychopathic Traits Inventory

The Youth Psychopathic Traits Inventory (YPI) is widely used in research, but there currently exist no means to identify potentially invalid protocols resulting from careless or random responding. We describe the development of an inconsistent responding scale for the YPI using three archival samples of youths, including two from the United States (juvenile justice and middle school) and one from Germany (vocational training school). We first identified pairs of correlated YPI items and then created a total score based on the sum of the absolute value of the differences for each item pair. The resulting scale strongly differentiated between genuine protocols and randomly generated YPI data (n = 1,000) across samples (AUC values = .88−.92). It also differentiated between genuine protocols and those same protocols after 50% of the original YPI items were replaced with random data (AUCs = .77−.84). Scores on this scale also demonstrated fairly consistent patterns of association with theoretically relevant correlates.

Learning Co-Substructures by Kernel Dependence Maximization

Modeling associations between items in a dataset is a problem that is frequently encountered in data and knowledge mining research. Most previous studies have simply applied a predefined fixed pattern for extracting the substructure of each item pair and then analyzed the associations between these substructures. Using such fixed patterns may not, however, capture the significant association. We, therefore, propose the novel machine learning task of extracting a strongly associated substructure pair (co-substructure) from each input item pair. We call this task dependent co-substructure extraction (DCSE), and formalize it as a dependence maximization problem. Then, we discuss critical issues with this task: the data sparsity problem and a huge search space. To address the data sparsity problem, we adopt the Hilbert--Schmidt independence criterion as an objective function. To improve search efficiency, we adopt the Metropolis--Hastings algorithm. We report the results of empirical evaluations, in which the proposed method is applied for acquiring and predicting narrative event pairs, an active task in the field of natural language processing.