讲座时间：6月21日
讲座地点：腾讯会议，会议号366-599-042
讲座（一）
讲座时间：14:00-14:45
讲座嘉宾：Ivan Blekanov
讲座主题：A Hybrid BERTopic-NetLogo Approaches for Evolutionary Simulation of Economic Information Flows in Social Media
讲座嘉宾简介
Ivan Blekanov，圣彼得堡国立大学应用数学系副教授。在科研与教育领域有多项优秀成果，曾获圣彼得堡市政府颁发的科学和教育工作奖。发表了超过90篇学术论文，部分收录在SCI等期刊中。在过去的五年里，其担任了十个信息技术与社会科学等跨学科项目团队负责人。与俄罗斯联邦天然气工业股份公司、华为及InterVeeam等公司保持着长期研发项目合作。此外，在2019年至2022年间其获得了八项专利（俄罗斯联邦知识产权局）。
讲座（二）
讲座时间：14:45-15:30
讲座嘉宾：Majid E. Abbasov
讲座主题：Rethinking a problem of obtaining cost optimal trajectory when planning transportation infrastructure
讲座嘉宾简介
Majid E. Abbasov，圣彼得堡国立大学教授，正博士。前后主持俄罗斯科学基金会项目3项，俄罗斯基础研究基金会1项目。
讲座摘要
The report addresses the issue of identifying the most cost-effective route between two fixed points on the terrain. Such issues typically arise in civil engineering as an important economic challenge when planning transportation infrastructure, as it is important to choose the best route in terms of construction cost. The standard approach nowadays relies on Geographic Information Systems (GIS) software, which is primarily graph-based. This involves creating a cost grid by dividing the area under consideration into square sub-areas using a uniform grid, with each sub-area assigned a price based on expert opinion or other assumptions. When constructing a route, it is therefore necessary to move exclusively along adjacent cells, from the cell containing the start point to the cell corresponding to the end point. This reduces the original problem to finding the shortest path on a graph connecting two nodes. Dijkstra's algorithm is widely used to solve this problem. However, increasing the density of the grid is necessary to obtain a more accurate solution, which dramatically increases the computational cost. To overcome this difficulty, researchers use heuristic modifications of Dijkstra's algorithm, such as the A* or D* search algorithms. However, these approaches can only produce suboptimal solutions and are therefore incapable of finding a solution with the required precision. Thus, the seemingly unsolvable problem of obtaining a solution with the required accuracy arises.
The report presents a new approach based on a novel mathematical formalisation. This enables researchers to solve the aforementioned problem and determine the optimal trajectory with the required level of accuracy. This involves results from optimisation, the calculus of variations, numerical methods, swarm intelligence algorithms, and other fields. This method of solving the problem has already demonstrated its effectiveness, and the research was supported by the Russian Science Foundation. While this project has successfully finished, there is still much to be done in this area. For example, the proposed approach could be strengthened and generalised by applying machine learning techniques. One could train a neural network to replicate the relationship between input parameters such as labour price, material expenditure, terrain restrictions, etc., and the optimal trajectory. This method, which was developed for mathematical physics problems, is called Deep Ritz, and it can be applied to the problem in question, too. All these topics will be discussed and presented in the report.
