Theses - Computation for Design and Optimization
http://hdl.handle.net/1721.1/39113
2014-09-21T16:22:57ZHigh-dimensional design space visualization for conceptual structural design
http://hdl.handle.net/1721.1/90083
High-dimensional design space visualization for conceptual structural design
Mueller, Caitlin T
This thesis focuses on visualizing high-dimensional design spaces for early-stage design problems in structural engineering and related disciplines. The design space, which is defined as the n + 1-dimensional surface that relates n design variables to a performance metric, contains all possible solutions to a formulated design problem. Graphical views of the design space are highly useful for designers because they organize a wide range of design possibilities in a compact, intuitive, and logical manner, illuminating global patterns, variable behaviors and relationships, and the nature of paths taken during iterative design processes. Design problems with two or fewer variables can easily be visualized in Euclidian space, through a curve or surface, but high-dimensional problems are difficult to display graphically. This is the key challenge addressed in this thesis. The thesis includes a critical review of existing methods for high-dimensional design space visualization, highlighting the unmet needs across a range of approaches. In response to these needs, the thesis makes a key contribution in the form of a new design space visualization method, called isoperforming parallel coordinate clusters (IPC clusters), that overcomes the issues of previous techniques. The IPC cluster approach is demonstrated on several conceptual structural design problems, and its application in optimization, directed exploration, and related design strategies is illustrated. Finally, the thesis concludes with a discussion of applications, impact, and future research directions. Key words: design space visualization, conceptual design, structural design, structural optimization
Thesis: S.M., Massachusetts Institute of Technology, Computation for Design and Optimization Program, 2014.; Cataloged from PDF version of thesis.; Includes bibliographical references (pages 63-66).
2014-01-01T00:00:00ZA thread-based parallel programming library for numerical algorithms
http://hdl.handle.net/1721.1/90080
A thread-based parallel programming library for numerical algorithms
Alhubail, Maitham Makki
This thesis presents a new simple lightweight C++ thread based parallelization library, intended for use in numerical algorithms. It provides simple multitasking and task synchronization functions. The library hides all internal system calls from the developer and utilizes thread pooling to provide better performance and utilization of system time and resources. The library is lightweight and platform independent, and has been tested on Linux, and Windows. Experiments were conducted to verify the proper functionality of the library and to show that parallelized algorithms on a single machine are more efficient than using the Message Passing Interface (MPI) using shared memory. In the opinion of several researchers who have used this library, the parallelized code is more easily understood and debugged than MPI. The results of initial experiments show that algorithms are as efficient or better than those using MPI.
Thesis: S.M., Massachusetts Institute of Technology, Computation for Design and Optimization Program, 2014.; Cataloged from PDF version of thesis.; Includes bibliographical references (page 47).
2014-01-01T00:00:00ZApproximation of the transient joint queue-length distribution in tandem networks
http://hdl.handle.net/1721.1/85470
Approximation of the transient joint queue-length distribution in tandem networks
Yamani, Jana H. (Jana Hashim)
This work considers an urban traffic network, and represents it as a Markovian queueing network. This work proposes an analytical approximation of the time-dependent joint queue-length distribution of the network. The challenge is to provide an accurate analytical description of between and within queue (i.e. link) dynamics, while deriving a tractable approach. In order to achieve this, we use an aggregate description of queue states (i.e. state space reduction). These are referred to as aggregate (queue-length) distributions. This reduces the dimensionality of the joint distribution. The proposed method is formulated over three different stages: we approximate the time-dependent aggregate distribution of 1) a single queue, 2) a tandem 3-queue network, 3) a tandem network of arbitrary size. The third stage decomposes the network into overlapping 3-queue sub-networks. The methods are validated versus simulation results. We then use the proposed tandem network model to solve an urban traffic signal control problem, and analyze the added value of accounting for time-dependent between queue dependency in traffic management problems for congested urban networks.
Thesis: S.M., Massachusetts Institute of Technology, Computation for Design and Optimization Program, 2013.; Cataloged from PDF version of thesis.; Includes bibliographical references (pages 95-97).
2013-01-01T00:00:00ZThe Markov chain Monte Carlo approach to importance sampling in stochastic programming
http://hdl.handle.net/1721.1/85220
The Markov chain Monte Carlo approach to importance sampling in stochastic programming
Ustun, Berk (Tevfik Berk)
Stochastic programming models are large-scale optimization problems that are used to facilitate decision-making under uncertainty. Optimization algorithms for such problems need to evaluate the expected future costs of current decisions, often referred to as the recourse function. In practice, this calculation is computationally difficult as it involves the evaluation of a multidimensional integral whose integrand is an optimization problem. Accordingly, the recourse function is estimated using quadrature rules or Monte Carlo methods. Although Monte Carlo methods present numerous computational benefits over quadrature rules, they require a large number of samples to produce accurate results when they are embedded in an optimization algorithm. We present an importance sampling framework for multistage stochastic programming that can produce accurate estimates of the recourse function using a fixed number of samples. Our framework uses Markov Chain Monte Carlo and Kernel Density Estimation algorithms to create a non-parametric importance sampling distribution that can form lower variance estimates of the recourse function. We demonstrate the increased accuracy and efficiency of our approach using numerical experiments in which we solve variants of the Newsvendor problem. Our results show that even a simple implementation of our framework produces highly accurate estimates of the optimal solution and optimal cost for stochastic programming models, especially those with increased variance, multimodal or rare-event distributions.
Thesis: S.M., Massachusetts Institute of Technology, Computation for Design and Optimization Program, 2012.; This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.; Cataloged from student-submitted PDF version of thesis.; Includes bibliographical references (pages 85-87).
2012-01-01T00:00:00Z