Novelty:
In modern society, the popular trend to effectively manage and operate repetitive tasks is knowledge transfer. People often do not do everything from the starting point, but apply knowledge gained from previous experiences to solve new problems. Obviously, we expect today’s intelligent systems to achieve similar capabilities. Knowledge transfer learning is a technique of reusing knowledge from solving one problem to solve another similar problem. This technique allows intelligent systems to find quality solutions to real-world problems in much less time. Currently, there are many forms of knowledge transfer used in algorithms: optimal knowledge transfer in evolution, knowledge transfer through co-evolution, and transfer in neural networks. Optimizing knowledge transfer in evolution provides knowledge from solving the previous problem to the evolutionary algorithm to solve the following problem. Knowledge transfer through co-evolution exploits the capabilities of many different optimization algorithms by designing a cooperation and information exchange mechanism between these algorithms. On the other hand, with transfer in neural networks, new neural networks are quickly trained using modifications from existing networks, thereby learning new recognition tasks almost immediately. This project focuses on improving knowledge transfer learning approaches for combinatorial optimization problems, with extensive applications in many fields of engineering and economics.
Impact:
This project has impacts in different aspects: applied theory, postgraduate training, developing a strong research team, connecting the research community. First, this project studies knowledge transfer methods for difficult optimization problems that will be applied to solve difficult NP problems in many urgent applications today such as scheduling, planning, transportation, network design, and a number of other artificial intelligence application fields. Next, the project also aims to publish top-notch research results of international stature, making important contributions to new technology solutions in the future. The project will build a strong research group on Evolutionary Computation and Machine Learning, connecting domestic and foreign scientists. Finally, the project contributes to training doctoral and master’s students to meet international standards.
