El Instituto de Computación (INCO) de la Facultad de Ingeniería de la Universidad de la República realizará un nuevo seminario académico titulado “The State of the art of neurodynamic optimization”, a cargo del investigador internacional Jun Wang. La actividad tendrá lugar el lunes 11 de mayo a las 15:00 horas en la Sala de Seminarios del INCO (salón 301).
El seminario abordará uno de los desafíos actuales más relevantes en el campo de la inteligencia artificial y la optimización computacional: cómo resolver problemas complejos de optimización en tiempo real mediante redes neuronales artificiales. La charla presentará avances recientes en modelos de optimización neurodinámica y sus aplicaciones en problemas científicos e ingenieriles de alta complejidad.
Abstract
As an important tool for scientific research and engineering applications, optimization is omnipresent in a wide variety of settings. It is computationally challenging to perform real-time optimization of dynamical systems. For such applications, classical optimization techniques may not be suitable due to the problem's high dimensionality and stringent time constraints. New paradigms are needed. One very promising approach to optimization is to apply artificial neural networks. Due to the inherent parallel and distributed nature of neural network information processing, the convergence rate of the solution process remains constant as the problem size increases. This talk will present the state of the art in neurodynamic optimization models and selected applications. Specifically, starting with the concept and motivation of neurodynamic optimization, I will review its historical background and present the state of the art, including various individual models for convex and generalized convex optimization. Additionally, I will present a multi-timescale neurodynamic approach to selected constrained optimization problems. In addition, I will introduce population-based collaborative neurodynamic approaches to constrained distributed and global optimization by deploying a population of individual neurodynamic models with diversified initial states at a lower level, coordinated by global search and information-exchange rules based on swarm intelligence at an upper level. Ultimately, I will demonstrate that many constrained optimization problems in science and engineering can be effectively and efficiently solved using neurodynamic optimization.
Sobre el expositor
Jun Wang es profesor catedrático de Inteligencia Computacional en el Departamento de Ciencias de la Computación y el Departamento de Ciencia de Datos de la City University of Hong Kong. Es una figura de referencia internacional en el área de redes neuronales e inteligencia artificial, con más de 350 artículos científicos publicados, editor de revistas de alto impacto como IEEE Transactions on Artificial Intelligence y distinguido con reconocimientos como el Neural Networks Pioneer Award y el Norbert Wiener Award. Además, es IEEE Fellow y miembro de diversas academias científicas internacionales.
Más información sobre los seminarios del INCO en: https://eva.fing.edu.uy/course/section.php?id=17045