Seminarios de Negocios 2015

Viernes 4 de diciembre, 11h | Sala 3 Alcorta

Juan José Miranda Bront (Universidad de Buenos Aires)

OPTIMIZACIÓN COMBINATORIA APLICADA A PROBLEMAS DE TRANSPORTE Y LOGÍSTICA

ABSTRACT

Los problemas de Optimización Combinatoria consisten en encontrar la solución óptima dentro de un conjunto grande (pero finito) de posibilidades. Estos surgen naturalmente en muchas situaciones reales, incluyendo la industria, marketing, robótica, logística y cualquier área relacionada con la toma de decisiones. En líneas generales, este tipo de problemas suele ser fácil de modelar matemáticamente pero difícil de resolver computacionalmente. El objetivo de este seminario es describir algunas de mis contribuciones realizadas en el desarrollo de métodos para problemas de transporte y logística.

En el área de Revenue Management, en los últimos años se han enriquecido los modelos tradicionales de demanda independiente mediante la incorporación del comportamiento selectivo de los clientes. En primer lugar, presentaré el desarrollo propuesto para la resolución de un modelo determinístico de programación lineal presente en la literatura, considerando un modelo de demanda más general donde los clientes pertenecen a segmentos superpuestos.

Luego, introduciré la investigación realizada en relación con problemas de logística y distribución, en particular sobre una generalización del conocido Travelling Salesman Problem (TSP) llamada Time-Dependent TSP (TDTSP), donde los tiempos de viaje no se asumen constantes. Este tipo de problemas ha recibido gran atención en los últimos años, motivados principalmente por obtener modelos más realistas para problemas de distribución en grandes ciudades. En este contexto, presentaré modelos de Programación Lineal Entera y los lineamientos generales de los algoritmos desarrollados para este problema.

Finalmente, comentaré brevemente algunas de las líneas actuales de investigación relacionadas a problemas tanto de planificación como operativos en la industria ferroviaria.

Jueves 19 de noviembre, 13:00hs | Lounge 1
Toon Devloo (Ghent University Belgium)

It’s not about the spark, but the flame: Unlocking the innovative potential of individuals and teams over time.

ABSTRACT

Organizations that aim to stimulate innovative work performance often come to the conclusion that it is very challenging for their employees to maintain or increase their initial motivation to innovate. This is, in part, due to the fact that most of the attention of so called ‘best practices’ goes to the spark that may initially encourage employees to innovate, but not to the development of motivation during long term innovation processes. In this research seminar, I will provide an overview of my PhD-research that focuses on the motivational underpinnings of creativity and innovation processes. More specifically, several studies will be presented that contribute, among other things, to a better understanding of how optimal forms of motivation are established to carry out innovative activities at the individual and team level. To do so, four fundamental challenges for current innovation research will be addressed.

First, the challenge to capture the dynamic nature of innovation processes across time.

Second, the challenge to develop new theoretical perspectives that deepen our knowledge on the motivational processes that drive creativity and innovation at a team-level.

Third, the challenge to account for the social realm in which creativity and innovation emerges and the identification of boundary conditions that affect the dynamic motivation-innovation relationship.

Finally, the challenge to gain a clearer sense of the unique motivational conditions that may stimulate exploration and exploitation activities which are both crucial for creative and innovative work.

Jueves 1 de octubre, 10:00hs. Sala Principal de Rectorado (Tercer Piso)
Enrico Gorgone (Université Libre de Bruxelles)

Convex optimization for large-scale problems: from Integer Programming to Machine Learning 

ABSTRACT

The focus of this presentation is to describe my research work that essentially concerns three principal topics: (i) Non-smooth Optimization, (iii) Heuristics and Combinatorial Optimization (ii) Machine Learning. The research in these fields is motivated by the industry's requirement to solve real world problems which are formulated as optimization problems. For this reason, in this presentation I describe the methodologies together with the applications and I present new algorithms designed for solving large-scale optimization problems with complex structure. Most of these problems are defined on networks, they occur in the context of public supply and commercial transportation, emergency response and risk management, industrial planning, etc. I discuss and compare several Lagrangian relaxation approaches that use techniques coming from the non-differentiable optimization like (incremental, deflected, projected) subgradient-type methods and (disaggregated, generalized) bundle type methods. In addition, I present an innovative subgradient scheme and a smart version of the bundle scheme suitable for the case of a sum-function. The numerical results of this comparison aim at providing a benchmark that might be used in other applications, pointing out the strengths and weaknesses of the different Lagrangian approaches. Finally, I present a branch and price method for the network design problem based on the Lagrangian relaxation.  I also present other real problems arising in medicine and in finance that I have tackled because of their socio-economic and societal impacts.