In order to promote multidisciplinary integration and interdisciplinary cooperation within the School, the School of Management and Engineering of Nanjing University held the third Intelligent Decision and Control Cross Forum of this semester on 20 November 2024 from 14:00-16:00 in Conference Room 204 of Xixin Building. Many teachers and students were invited to attend this event. The forum was hosted by Mr Wang Zhengli, and Mr Li Huaxiong from the Department of Control Science and Intelligent Engineering and Mr Hu Qian from the Department of Industrial Engineering and Operations Management shared their latest research results, and the presentations of the two teachers were also the 19th and 20th lectures of this forum.

Mr Li Huaxiong from Department of Control Science and Intelligent Engineering gave a lecture entitled ‘Optimisation Methods in Multimodal Hash Learning and Multiview Learning’. The presentation introduced one of the important directions in the field of machine learning, i.e. multimodal hash learning and multi-view learning. The models can be modelled as optimisation problems under constraints and solved by optimisation methods. Usually, these optimisation models are difficult to find closed-form solutions and need to be solved numerically by using alternating optimisation methods. The alternating direction multiplier method is an important alternating optimisation method under constraints, which decomposes the optimisation objective function into more easily solvable sub-objective functions, and iterates the sub-objective functions by alternating updates to find the optimal solution of the original objective function. This report introduces the alternating optimisation method and its application in multimodal hash learning and multiview learning from various perspectives, such as adaptive label-associated cross-modal hash learning and feature decoupling-based multiview learning.

Mr Hu Qian from Department of Industrial Engineering and Operations Management gave a lecture entitled ‘Modelling and Algorithmic Research on Network Capacity Expansion Problem’. The presentation introduced that network design has important applications in telecommunication, transport, manufacturing and other industries. With the rapid growth of business scale, how to expand the capacity based on the existing network has become an important management decision problem. Based on the business scenarios of telecommunication networks and considering the constraints of whether the traffic is divisible or not, a class of multi-commodity flow network scaling problems is investigated, i.e., the scaling and traffic allocation decisions are made to minimise the total cost while satisfying the constraints of the traffic demand of different services, the capacity limitations, and the traffic splitting limitations. Based on the analysis of the problem structure, arc flow-based and path flow-based mathematical models are established respectively, and the exact algorithms for solving the problem are designed. Focusing on the fast solution algorithms for key subproblems, effective inequalities and fast separation and lower bound improvement, local search and upper bound improvement based on flow transfer, these key designs make the algorithms run efficiently and are verified in computational experiments. The research results help to improve the management decision of improving service network design and capacity expansion and upgrading planning, and provide efficient algorithmic technical support.

Participating teachers and students asked questions and exchanged views on the reports of the above two teachers. The forum was successfully concluded.