UROP Proceedings 2020-21

School of Engineering Department of Computer Science and Engineering 145 Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: PATUPAT Albert John Lalim / DSCT Course: UROP2100, Spring Recommender systems have been serving as increasingly important navigation tools tackling information overload present in the digital economy. As an emerging paradigm of this field, anonymous session-based recommender systems model short-term yet dynamic user preferences solely based on anonymous sessions and interactions. This report reviews the different approaches for anonymous session-based recommendation, highlighting representative algorithms. Afterwards, several weaknesses with current methods are discussed; most notably, readily available timestamps of user actions are often discarded despite carrying meaningful information. Ultimately, this paper proposes a timestamp-aware graph neural network addressing these problems. Future direction of such work includes empirical analysis of the proposed model, theoretical analysis of the training and prediction algorithm, and extension of the approach for online data stream environments. Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LAU Yan Hei / SENG Course: UROP1000, Summer The following report consists of a summary of the knowledge acquired during the UROP1000 course, under the guidance of the instructor. The courses studied include COMP2011 Introduction to Object-oriented Programming, COMP2012 Object-Oriented Programming and Data Structures and COMP3711 Design and Analysis of Algorithms. The contents below serve to demonstrate that the student has achieved a thorough understanding of the materials presented in the above courses. Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LIU Zichen / DSCT Course: UROP1000, Summer Collective Spatial Keyword Queries (CoSKQ) aims to find a set of spatial objects that collectively satisfy keywords requirements. However, the queries studied so far generally focusing on “covering” the requirement. Among these query keywords, some of them may be of particular interest. We define the Interest-Aware and Distance-Constrained Collective Spatial Keyword Query such that the returning set not only covers the keywords but also carries as many objects of interest as possible under a distance threshold. Specifically, we study 3 variants of this problem based on different distance functions for the set. In this paper, we prove that it is NP-hard to solve the query with any constant factor and provide both exact and bi-approximation algorithms.