Abstract

Variational autoencoders were proven successful in domains such as computer vision and speech processing. Their adoption for modeling user preferences is still unexplored, although recently it is starting to gain attention in the current literature. In this work, we propose a model which extends variational autoencoders by exploiting the rich information present in the past preference history. We introduce a recurrent version of the VAE, where instead of passing a subset of the whole history regardless of temporal dependencies, we rather pass the consumption sequence subset through a recurrent neural network. At each time-step of the RNN, the sequence is fed through a series of fully-connected layers, the output of which models the probability distribution of the most likely future preferences. We show that handling temporal information is crucial for improving the accuracy of the VAE: In fact, our model beats the current state-of-the-art by valuable margins because of its ability to capture temporal dependencies among the user-consumption sequence using the recurrent encoder still keeping the fundamentals of variational autoencoders intact.

Abstract

The Interactions website (interactions.acm.org) hosts a stable of bloggers who share insights and observations on HCI, often challenging current practices. Each issue we'll publish selected posts from some of the leading and emerging voices in the field.

Abstract

The disclosed embodiments illustrate method and system for auto-generation of the sketch notes-based visual summary of the multimedia content. The method includes determining one or more segments, based on one or more transitions in the multimedia content. The method further includes generating a transcript based on audio content associated with each determined segment. The method further includes retrieving a set of images from an image repository based on each of the identified one or more keywords. The method further includes generating a sketch image of each of one or more of the retrieved set of images associated with each of the identified one or more keywords. The method further includes rendering the sketch notes-based visual summary of the multimedia content, generated based on at least generated one or more sketch images, on a user interface displayed on a display screen of the user …

Abstract

India is home to the largest under-25 demographic profile in the world, but lacks a job-ready educational system. It requires a wide-spread, skill-oriented educational model, equipping youth to thrive in highly dynamic job markets. As a response to the huge demand for technical education, a large private skill-tutoring ecosystem has sprung up in In-dia but remains geographically limited. This paper, drawn from a three-month ethnographic research conducted in Ameerpet (arguably India's largest IT skilling hub), probes the pedagogic style and characteristics of tutoring, and of-fers reasons why learners prefer to enroll into a physical model of classroom teaching over online courses. We make design suggestions for online learning platforms to attract students who are marginalized in the more formal and com-petitive education system, and opt for Ameerpet-like skill-hubs. Our primary offering is to suggest a shift in …

Abstract

The disclosed embodiments illustrate methods and systems to recommend medical intervention activities for caregivers and/or human subjects. The method comprises receipt of medical data of a human subject and contextual data of a plurality of caregivers of the human subjects. The medical data includes periodic physiological data of the human subject received in real-time. A customized ontology is generated, from a pre-determined ontology related to a medical condition, based on the received medical data and the contextual data. Thereafter, a recommendation of medical intervention activities is generated for at least one of the plurality of caregivers and/or the human subject, based on analysis of the generated customized ontology, caregiver profiles, the received contextual data, and the periodic physiological data. The generated recommendation is transmitted to computing devices of the at least one …

Abstract

In this work, we consider the problem of decentralized multi-robot target tracking and obstacle avoidance in dynamic environments. Each robot executes a local motion planning algorithm which is based on model predictive control (MPC). The planner is designed as a quadratic program, subject to constraints on robot dynamics and obstacle avoidance. Repulsive potential field functions are employed to avoid obstacles. The novelty of our approach lies in embedding these non-linear potential field functions as constraints within a convex optimization framework. Our method convexifies nonconvex constraints and dependencies, by replacing them as pre-computed external input forces in robot dynamics. The proposed algorithm additionally incorporates different methods to avoid field local minima problems associated with using potential field functions in planning. The motion planner does not enforce predefined trajectories or any formation geometry on the robots and is a comprehensive solution for cooperative obstacle avoidance in the context of multi-robot target tracking. We perform simulation studies for different scenarios to showcase the convergence and efficacy of the proposed algorithm.

Abstract

Efficacy of games aslearning medium is of interest to researchers and the gaming industry. A critical metric for learning is knowledge retention and very few studies have conducted in-depth comparisons of: a) game versus no-game learning, b) collaborative versus individual learning. Towards this, the study reported in this article will present the findings from an experiment using Asia and world maps, cut into pieces as in a jigsaw puzzle. The participants were primary school children who were randomly assigned to the puzzle and no-puzzle group. To understand the role of collaborative interactions in learning, each group was further divided into two subgroups. Each subgroup either solved the puzzle or studied the full map (no-puzzle) individually or collaboratively. Three post-tests were conducted over a period of 10 days. The mean scores and Mann Whitney test shows: a) In the no-puzzle condition, no difference in the average scores of the individual and collaborative groups for both maps was observed, b) In the puzzle condition, the collaborative group score was slightly more than that of the individual group for Asia map, while the difference was significant for the continent map, and c) Puzzle and no-puzzle individual group scores were comparative for Asia map but the continent map groups showed a major difference. The findings are mixed with collaborative puzzle solving showing higher retention while puzzle solving does not show significant effect on learning and retention.

Abstract

A professional player's expertise rests on the ability to predict action by optimally extracting the opponent's postural cues. Eye tracking (head-mounted system) data in a naturalistic singles badminton play was collected from one professional wor0ld-ranked player facing five amateur players (10 serves or 50 trials) and two amateurs playing against four other amateur players each (10 serves or 80 trials). The visual gaze on the opponent body, segregated into 3 areas-of-interest covering the feet, face/torso, and hand/racket of the opponent and the shuttle, was analysed for a) the period just before the serve, b) while receiving the serve and c) the entire rally. The comparative analysis shows the first area-of-interest for professional player as the opponent's feet while executing the serve and the hand/racket when receiving a serve. On the other hand, the amateur players show no particular strategy of fixation location either for the serve task or while facing a serve. The average fixation duration (just before serve) for the professional was 0.96s and for the amateurs it was 1.48s. The findings highlight the differences in the postural cue considered important and the preparatory time in professional and amateur players. We believe, analytical models from dynamic gaze behavior in naturalistic game conditions as applied in this study can be used for enhancing perceptual-cognitive skills during training.

Abstract

Weighted Frequent Itemset (WFI) mining is an important model in data mining. The popular adoption and successful industrial application of this model has been hindered by the following two obstacles: (i) finding WFIs is a computationally expensiveness process as these itemsets do not satisfy the downward closure property and (ii) lack of parallel algorithms to find WFIs in very large databases (e.g. astronomical data and twitter data). This paper makes an effort to address these two obstacles. Two pattern-growth algorithms, Sequential Weighted Frequent Pattern-growth and Parallel Weighted Frequent Pattern-growth, have been introduced to discover WFIs efficiently. Both algorithms employ three novel pruning techniques to reduce the computational cost effectively. The first pruning technique prunes some of the uninteresting items by employing a criterion known as cutoff weight. The second pruning technique, called conditional pattern base elimination, eliminates the construction of conditional pattern bases if a suffix item is an uninteresting item. The third pruning technique, called pattern-growth termination, defines a new terminating condition for the pattern-growth technique. Experimental results demonstrate that the proposed algorithms are memory and runtime efficient, and highly scalable as well.

Abstract

Efficient discovery of interesting patterns using parallel algorithms is an actively studied topic in data mining. A key research issue related to this topic is data segmentation, which influences the overall computational requirements of an algorithm. This paper makes an effort to address this issue in correlated pattern mining. Two novel data segmentation techniques, ‘database segmentation’ and ‘transaction segmentation,’ have been introduced to discover the patterns efficiently. The former technique involves segmenting the database into multiple sub-databases such that each sub-database can be mined independently. The latter technique involves segmenting a transaction into multiple sub-transactions such that each sub-transaction can be processed as an individual transaction. The proposed techniques are algorithm independent, and therefore, can be incorporated into any parallel algorithm to find correlated patterns effectively. In this paper, we introduce map-reduce based pattern-growth algorithm by incorporating the above mentioned techniques. Experimental results demonstrate that the proposed algorithm is memory and runtime efficient and highly scalable as well.