School of Engineering Department of Computer Science and Engineering 124 Efficient Queries Over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LIAO Junyu / COSC Course: UROP 1100, Summer This report presents the current progress of my undergraduate research opportunities program, which mainly encompasses the completion of two courses, respectively, COMP2012 Object-Oriented Programming and Data Structures, and COMP3711 Design and Analysis of Algorithms, before actually starting the research work. The completion of the two courses provided a robust foundation in programming principles, data structures, and algorithmic techniques. The concept of the two courses is of great importance in the field of computer science and is critical to successful conducting the following research work of the UROP project. This report detailly concludes the concepts I learned in the two courses, as well as a broader understanding of conducting research. Efficient Queries Over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LUI Ka Kit / COMP Course: UROP 1100, Fall UROP 2100, Spring In order to generate useful Data visualizations (DVs) for knowledge discovery in dataset, declarative visualization languages (DVLs) such as Vega-Lite, ggplot2 are used. However, it could be difficult to learn these DVLs, especially to non-technical users. To make this more approachable, one may use natural language queries to specify the DVs, transform them into the DVLs, and use them to generate visualizations. This transformation is known as NL2VIS problem. This report proposes the use of generative large language models (LLMs) to tackle the NL2VIS problem. By fine-tuning the model Vicuna-7B, we expect the performance will be better than existing approaches. We further extended our work in the previous semester and performed some experiments to investigate the use of the LLMs. Efficient Queries Over Database Supervisor: WONG Raymond Chi Wing / CSE Student: ZHANG Jinming / DSCT Course: UROP 3200, Summer In the previous semester, a new session recommendation system EDFNN which extends the GNN to continuous space was proposed. Further optimization of the proposed model was performed including performance improvement, theoretical supports, and experiments with various baselines under different environments. This report gives a more detailed analysis of the target problems, related studies, and how our proposed model is better than existing models. Experiment details and discoveries are also elaborated with a potential new architecture.
RkJQdWJsaXNoZXIy NDk5Njg=