Noticias en español
Mathematica for Teaching and Education
Abstract: Learn how to improve your classroom experience with Mathematica. This course gives a tour of functionality relevant to teaching and learning, along with case studies and best-practice suggestions for course integration. Topics include making your classroom dynamic with interactive models and a survey of computation and visualization capabilities useful for teaching practically any subject at any level.
Read MoreMPC Motivited Computational Optimization
Abstract: I will start with a brief introduction to the paradigm of Model Predictive Control (MPC). Then I will discuss various enhancements of the Newton/SQP method motivited by MPC problems, such as modified and inexact Newton-type methods. Finally, I will talk about tracking solution trajectoris of parametric variational inequalities by using predictor-corrector path-following.
Read MoreComplexity in Convex Optimization: Oracles, Algorithms and Applications
Abastract The need to solve large-scale optimization problems is ubiquitous. Modern applications in signal processing, machine learning and big data, need at their core conv optimization solvers. In this respect, understanding the limits of performance of such methods is a crucial task. In this talk we introduce the theory of oracle complexity as a standpoint to understand the efficiency of optimization algorithms. We will see how state of the art algorithms can be seen as oracle-based, and how the classical theory does not address current applications. Motivated by the later...
Read MoreImplicaciones geométricas de la desigualdad de Poincaré en espacios métricos de medida
Resumen: En los últimos años, el cálculo de primer orden desarrollado clasicamente en el marco de los espacios euclídeos, se ha extendido a espacios que no necesariamente están dotados de una estructura diferenciable. Una vertiente del cálculo de primer orden que se ha desarrollado en el contexto general de espacios métricos ha sido bajo la hipótesis de que nuestra espacio admita una medida doblante y una desigualdad de Poincaré. Dicha desigualdad crea una conexión entre la métrica, la medida y el módulo del gradiente, además de un nexo de union entre el comportamiento global y local de las...
Read MoreUn enfoque mixto para la optimización de políticas energéticas
RESUMEN: Existen variados enfoques clasicos utilizados para la optimizacion de politicas energeticas. Cada uno de ellos supone ciertos requerimientos, tales como la convexidad de las funciones de costo o restricciones sobre los procesos estocasticos, los cuales pueden resultar beneficiosos para el analisis teorico o de simulaciones pero perjudicuales a la hora de aplicar estos metodos en la vida real. Para enfrentar este problema, proponemos un enfoque mixto que incorpora la programacion dinamica y metodos de aprendizaje automatico: la Busqueda Directa de Valores (Direct Value Search DPS)....
Read MoreStochastic dynamic programming applied to microgrid management
Abstract: Energy Management Systems (EMS) are usually based on the integer programming framework. A novel method based on stochastic control and dynamic programming is proposed. It allows for a precise modeling of complex constraints such as the minimal working power of a diesel motor. The EMS proposed here is based on a rolling horizon strategy. It takes into account historical data and randomness as well as the aging of the battery and the motor working constraints. A semi-lagrangian scheme was implemented within BOCOP – a software developed at INRIA – for the numerical...
Read More