📅 2 июня в 11:00 (мск) | Онлайн
Тема: Learning-Based Algorithms for Solving Combinatorial Optimization
This talk explores learning-based optimization paradigms that deeply integrates data, models, and algorithms for combinatorial optimization. For binary integer programming, a Monte Carlo policy gradient method is employed to achieve efficient sampling and search within the feasible region. For the quadratic assignment problem, an integrated framework of offline training and online fine-tuning is constructed, balancing solution accuracy and scalability. In complex constrained path planning and directed acyclic graph-based heterogeneous scheduling, lazy mask decoding and heuristic search are synergistically combined. Extensive experiments demonstrate that the learning-based algorithms exhibit significant advantages in solution quality, feasibility, and robustness.
Спикер: Zaiwen Wen — Boya Distinguished Professor в Пекинском университете (PKU)
Специалист в области алгоритмов оптимизации, машинного обучения и ИИ. Лауреат China Youth Science and Technology Award (2016), член редколлегий ведущих международных журналов, вице-президент Operations Research Society of China.
📆 Добавить семинар в ваш календарь
🔗 Ссылка для подключения