Dynamic Optimization

Organized and chaired by Dr. Emilia TantarDr. Alexandru-Adrian Tantar and Dr. Peter Bosman.

 

 

tantar-emilia

Dr. Emilia Tantar, University of Luxembourg, Luxembourg

She received both her Diploma degree and MsC from the Computer Science Faculty at the "Al. I. Cuza University" in Iasi, Romania. In 2005 she joined the French National Institute for Research in Computer Science and Control(INRIA) in Lille. She was awarded the PhD title for Landscape analysis in multi-objective optimization in 2009 at the University of Lille 1. Between 2007 and 2009 she hold a lecturer position at the same university. During her PhD she was also awarded an INRIA Explorateurs grant to the CWI, Amsterdam, Netherlands. She developed a strong interest on new challenging aspects regarding landscape analysis in multi-objective optimization, but also the theoretical foundations of stochastic methods and their scaling to practical problems. Before joining the CSC research unit as a Marie Curie (AFR grant) researcher, at the University of Luxembourg, in October 2010, she was an INRIA post-doctoral researcher in the Advanced Learning Evolutionary Algorithms (ALEA) team, at INRIA Bordeaux, dealing with performance guarantees factors for multi-objective particle methods, such as evolutionary algorithms and rare event simulation techniques. Emilia is currently co-authoring with Oliver Schutze a book in Springer series, dealing with performance guarantees and landscape analysis in multi-objective optimization and also co-chairing since 2011 the GreenGEC workshop at GECCO.

 



tantar-alexandru

Dr. Alexandru-Adrian Tantar, University of Luxembourg, Luxembourg

Dr.  Alexandru-Adrian Tantar received his Ph.D. diploma in Computer Science in 2009 from the University of Lille. He was a member of the DOLPHIN team, INRIA Lille - Nord Europe / LIFL (French National Institute for Research in Computer Science and Control / Fundamental Computer Science Laboratory of Lille). By the end of 2008 he was on a short stay in the SEN4 team, CWI, Amsterdam, The Netherlands (INRIA Explorateurs Grant, September 2008 - December 2008). As a postdoctoral researcher in the Advanced Learning Evolutionary Algorithms team, INRIA Bordeaux - Sud-Ouest, he worked on parallel interacting Markov chains based algorithms (September 2009 - March 2010), he co-organized the ALEA working group and the "Rare Events Simulation 2010" workshop. He worked with the Atomic Energy Commission (Life Sciences Division and CESTA), the Biology Institute of Lille and the Sea French Research Institute. Since the 1st of April 2010, Dr. Tantar is a Marie Curie (AFR Grant) postdoctoral researcher in the Computer Science and Communications Research Unit, University of Luxembourg. He is involved in the GreenIT (energy-efficient solutions for Cloud Computing / HPC centers, FNR Core 2010-2012). He also co-organizes the GRIPHON working group (CSC), he participates to the University of Luxembourg's Carbon Neutral ICT Operations program. He is co-chairing since 2011 the GreenGEC workshop at GECCO and provide a tutorial on Green Evolutionary Computing for Sustainable Environments at CEC 2012.


bosman-p

Dr. Peter Bosman, Centrum Wiskunde und Informatica, Netherlands

Peter A.N. Bosman is a scientific staff member in the research group Multi-agent and Adaptive Computation at the Centrum voor Wiskunde en Informatica (CWI) (Centre for Mathematics and Computer Science) located in Amsterdam, the Netherlands. Peter was formerly affiliated with the Department of Information and Computing Sciences at Utrecht University, where also heobtained both his MSc and PhD degrees in Computer Science, more specifically on the design and application of a specific type of evolutionary algorithm, the estimation-of-distribution algorithm (EDA). He has since then been an active researcher in the field of evolutionary computation. His current research position is mainly focused on fundamental EDA research and on applications of EDAs and (multi-)agent technology in energy systems, revenue management and the life sciences. Peter is best known for status of active researcher in the area of EDAs since its upcoming within the area of Genetic and Evolutionary Computation (GEC) and has (co-)authored some 40 publications in the GEC area. At the GECCO conference, Dr. Bosman has previously been track chair (EDA track, 2006, 2009), late-breaking-papers chair(2007) and co-workshop chair.

 



 
 

In the management of critical infrastructures that our society is based upon, often on-line (i.e. real-time) solutions are required. The associated optimization problems are dynamically changing, meaning that the quality of candidate solutions may be different at different points in time. Many real-life processes and problems are intrinsically dynamic. Typical examples can be found in logistics, e.g. routing a fleet of vehicles where the loads can be announced in real-time or managing a supply chain of goods, but also in other areas such as biology, e.g. protein interactions in a living body (classically modeled in a static form), anthropology and the (dynamic) evolution of species and in physics such as magnetic interactions and spin structures.

Optimization problems may be dynamic for different reasons. A special case is uncertainty, or noise, which is often modeled as stochastic perturbations of the true objective function. Another source of dynamism comes from the fact that environment in itself may be subject to external factors that evolve with time, such as the impact that weather changes have on renewable energy supplies for green technologies. A third source of dynamism is given by state dependency or time-dependent constraints as typically found in feedback and decision-based systems and is often studied in reinforcement learning.

This session firstly aims at providing a thorough understanding of what dynamic problems stand for. The session further aims at offering an overview of the existing paradigms and their adaptation for real-life dynamic problems. A particular interest, besides tracking optima (or sets of best compromise solutions in case the problem is multi-objective), is taken for learning, anticipation and resilience to extreme noise or severe changes. Ways to measure the quality of a solution when dealing with dynamic problems (performance measures and comparison) is also of special interest. As an ultimate aim, this session is intended to bring together researchers from different fields that are interested in better understanding dynamic optimization problems and approaches to solve such problems. The session is open to researchers from all relevant fields and welcomes both academic and industrial applications, including (but not limited to) the following topics: 

  •  Dynamic systems
  •  Evolutionary and probabilistic dynamic approaches 
  •  Dynamic single- and multi-objective optimization 
  •  Learning and anticipation  
  •  Handling of uncertainty at different levels (fitness function, variables, environment, constraints) 
  •  Self-adaptive methods for dynamic environments 
  •  Self-adaptive representations for dynamic optimization problems 
  •  Cross-domain dynamic paradigms 
  •  Theoretical models issued from physics, bioinformatics or other fields where dynamics are intrinsic 
  •  Platforms for dynamic single or multi-objective optimization
  •  Real-life applications 
  •  Robustness of the provided solutions 
  •  Performance measures, analysis and comparison of algorithms or results
  •  Statistical testing and comparison of algorithm performances
  •  Theoretical analysis
In