School of Engineering Department of Computer Science and Engineering 127 Knowledge Discovery Over Database Supervisor: WONG Raymond Chi Wing / CSE Student: YIP Sau Lai / DSCT Course: UROP 1100, Summer This progress report details my research under the UROP1100 project titled “Knowledge Discovery Over Database.” My activities during this summer session are divided into two primary stages: (1) a comprehensive exploration of the field of databases, and (2) preparations for developing a VQL debugger. As a novice in database research, I dedicated the first half of the session to acquiring a broad understanding of the field by studying various sub-areas extensively. I reviewed suggested papers on two main streams: algorithm-based and model-based problems. Based on my interest and comprehension, I decided to work on developing a debugger for model-generated VQL queries. The latter half was spent preparing the pipeline design by reviewing related work on SQL and other programming languages. Knowledge Discovery Over Database Supervisor: WONG Raymond Chi Wing / CSE Student: ZHANG Jinming / DSCT Course: UROP 1100, Spring Session recommendation systems are crucial for social medias and online shopping platforms to make accurate recommendations when the user profile is not available. Several models and algorithms have been proposed, while GNN-based models show better performance than the others for their efficiency in extracting inter-item relationships on the session graphs. We have proposed edge density function (EDFNN) as an extension of GNN in continuous node spaces for session recommendations and verified its performance on baseline models such as SRGNN and LESSR. However, there are more details about our proposed model that need to be carefully elaborated, such as its over-smoothing and reliability of the NICE model used as embedding function. This report provides supplementary information about EDFNN with theoretical analysis regarding the aforementioned issues and potential improvement about the model in future works. Knowledge Discovery Over Database Supervisor: WONG Raymond Chi Wing / CSE Student: ZHANG Zongmin / COMP Course: UROP 1100, Summer The popularity of recommendation systems is steadily increasing, with various types being widely employed in E-commerce and video websites. Many companies, such as Amazon or Netflix, utilize these systems to predict user actions and provide personalized recommendations. By employing session-based recommendation with graph neural networks, the intricate transitions between items can be more effectively captured, leading to more precise recommendations for the users. This UROP project aims to comprehensively study the fundamental concepts of recommendation systems while understanding the principles behind different algorithms and data filtering methods. Additionally, we will develop tools that optimize computational efficiency during training.
RkJQdWJsaXNoZXIy NDk5Njg=