糖心TV

PhD Thesis title (Complexity Sciences and Operational Research): Efficient sequential sampling for global optimization in static and dynamic environments.

Many optimization problems involve acquiring information about the underlying process to be optimized in order to identify promising solutions. Moreover, in some cases obtaining this information can be expensive, which calls for a method capable of predicting promising solutions so that the global optimum can be found with as few function evaluations as possible. Another kind of optimization problem arises when dealing with objective functions that change over time, which requires tracking of the global optima over time. However, tracking usually has to be quick, which excludes re-optimization from scratch every time the problem changes. Instead, it is important to make good use of the history of the search even after the environment has changed.

My research revolves around the topic of response surface based sequential sampling for global optimization of expensive-to-evaluate black-box functions under static and dynamic scenarios.
Regarding the former scenario, it addresses the high computational cost inherent to Efficient Global Optimization (EGO), a global search algorithm that is known to work well for expensive black-box optimization problems where only few function evaluations are possible, and which uses surrogate models of the fitness landscape for deciding where to sample next. The proposed variant is based on partitioning the space and building local models to accelerate the selection of future sampling locations with a minimal impact on the optimization performance. The linear computational complexity as a function of the number of observations of this extension is shown, and its performance benchmarked to both the original algorithm it extends, and state of the art algorithms.
For the latter scenario, we propose and compare four methods of incorporating old and recent information in the surrogate models of EGO in order to accelerate the search for the global optima in a dynamically changing environment. As we demonstrate, exploiting old information as much as possible significantly improves the tracking behavior of the algorithm.

Furthermore, using similar techniques, I explore how to incorporate information coming from models of the same system but with different fidelities. This is useful when dealing with expensive but very accurate simulations, and cheap but less accurate ones.

Supervisor:

Co-Supervisor:

Research Interests:

• Statistics and Machine Learning
• Stochastic Optimization
• Dynamic Optimization
  
• Approximate Dynamic Programming
• Simulation
• Revenue Management
• Optimal Dynamic Pricing
• Evolutionary Algorithms and Heuristics
• Gaussian Processes
• Time series clustering

Publications and talks

• Paper: Morales-Enciso, Sergio, and Branke, Juergen. Tracking global optima in dynamic environments with efficient global optimization (Submitted).
  
• Paper: Morales-Enciso, Sergio, and Branke, Juergen. Response Surfaces with Discounted Information for Global Optima Tracking in Dynamic Environments. To appear in Studies in Computational Intelligence, Springer, 2013.
• Conference talk: Morales-Enciso, Sergio, Response Surfaces with Discounted Information for Global Optima Tracking in Dynamic Environments. (, Canterbury, UK).
• Conference talk: Morales-Enciso, Sergio, (What is Complexity Science?) at The European Parliament, Strasbourg, France, November 2012. ()
  
• Paper: Morales-Enciso, Sergio, and Branke, Juergen. In Dagstuhl OpenAccess Series in Informatics (OASIcs).
• Conference talk: Morales-Enciso, Sergio, Sampling policies for revenue maximization through dynamic pricing under unknown market behaviour. (, Nottingham).
• Conference talk: Morales-Enciso, Sergio, Best next sample. (, Bristol).
• Industrial collaboration: Bandeira, A., Cavaleiro, M., Davies, R., Mondal, A., Morales, S., Piwarska, K, Please, C. European Study Group with Industry 73, University of 糖心TV, 2010.

M.Sc. Dissertations :

• Due Date Prediction in Self-Organized Manufacturing Plants
  • Abstract: [Five due date prediction models are compared in a simulated self-organized manufacturing plant working under the critical ratio dispatching rule. The problem of generating training data from an uninformed state when using a due date dependent dispatching rule is approached by using due date setting methods not requiring any prior information of the system. Several replicas of the numerical experiments are performed and the results obtained are tested for statistical significance. A set of Pareto-optimal configurations is proposed according to the accuracy and the efficiency achieved by the compared due date setting and prediction models.]
  • Supervisor:

• Parameter Estimation for a Multi-Agent Driven Crowd Movement Model
  • Abstract: [The current work addresses the use of a genetic algorithm to find a best fit for non measurable parameters in an adaptable crowd movement model. The parameters must be inferred from high resolution data obtained from experiments aiming to track human movement in real non panic generic situations. The methodology proposed provides the means to create real size models of complete systems based on small scale experiments. The mathematical description of the model is provided as well as the fundamental guidelines for its practical implementation. The proposed methodology is validated with numerical experiments proving the flexibility of the model to adapt to arbitrary scenarios and providing best fit estimators for the non measurable parameters along with their confidence interval.]
  • Poster:
  • Supervisor:

 

Co-supervised master thesis:

• 2011 - Understanding the Price for Fast Delivery in a Build-to-Order
  Manufacturing Firm. Samuel Sze Ming Chan (WBS)

Independently Organised Workshops

• November 2012, The University of 糖心TV, England.

Teacher assistantships:

• Spring 2013 - IB98D0 Advanced Data Analysis (Masters level). (糖心TV 糖心TV School)
  
• Spring 2013 - Quantitative Analysis for Management II. (糖心TV 糖心TV School)
• Fall 2012 - Quantitative Analysis for Management I. (糖心TV 糖心TV School)
• Spring 2012 - Quantitative Analysis for Management II. (糖心TV 糖心TV School)

Background:

• MSc in Complexity Sciences at
• MEng in Information Technologies (minor in Strategic Planning and Management) at
• BEng in Information Technologies and Telecommunications at

Selected attended workshops and conferences:

• , November 2012. London.
  
• , June 2012. Madrid Stock Exchange, Spain.
  
• , April 2012. Imperial College London.
• 
  

  • Heuristics & Approximation Algorithms, April 2012. The University of Nottingham.
    

  • Simulation, July 2011. The University of 糖心TV.
    
  • Systems Dynamics, July 2011. The University of 糖心TV.
    
• training courses:
  • , September 2011. The University of 糖心TV.
  • , December 2010. The University of 糖心TV.
  • , November 2010. The University of 糖心TV.
• , October 2010. London.
  

Complexity Science Complex
D2.05, Zeeman Building
University of 糖心TV
Coventry, CV4 7AL, UK

S.Morales-Enciso@warwick.ac.uk