Dynamic resource management and numerical optimization : practical applications in communication networks and finance
- The exponential increase in computing efficiency over the past few decades, known as Moore's law, has led to wide-spread commercial applications of resource management and optimization theory in various fields. This thesis researches the increasingly important roles dynamic and intelligent resource management and optimization algorithms have played in present-day communication networks and financial portfolio management. First, this thesis studies Dynamical Spectrum Management (DSM), an emerging technology for next-generation Digital Subscriber Line (DSL) access networks. Present DSL networks' speed and Quality-of Service (QoS) are often limited by the interference between multiple users sharing a common physical transmission medium. Compared to the traditional static network planning that limits network speed to survive laboratory contrived worst-case scenarios, DSM's dynamic resource management and optimization adapts to network channel and noise changes to improve speed and robustness significantly. A fast and robust Multi-Level Water-Filling (MLWF) DSM algorithm is proposed for the efficient management of multi-user DSL networks. Distributed protocols are also proposed for MLWF's low-overhead implementation in DSL access networks. Second, this thesis shows that DSM can also benefit Home Power-line Communication (PLC) networks, which utilize home power cords as the transmission medium with a Frequency-Division Multiple-Access (FDMA) resource allocation constraint. An Equivalent Interference (EI) is introduced to reformulate the nonlinear integer programming Home-PLC FDMA resource management problem into a special case of DSL's multi-user interference management problem. DSM algorithms can thus be applied to solve the FDMA resource allocation problem in Home-PLC networks. In addition, dynamic relay protocols based on the presented FDMA algorithms are proposed to reuse idle Home-PLC devices to increase PLC networks' data rate substantially. Third, this thesis studies dynamic asset allocation strategies for drawdown-limited portfolio management. The divergence between theoretical model assumptions and empirical financial time-series statistics are analyzed. Empirical simulation results demonstrate that theoretical drawdown-limited portfolio management strategies tend to over-leverage a portfolio with lagged and overly optimistic correlation and volatility estimate. Adjustment on the trading strategy is thus needed to balance the risk and return of a portfolio. Finally, this thesis analyzes investment managers' incentive alignment with investors under a utility maximization framework. Simulation results illustrate that mangers' and investors' interests can often be misaligned. An investor-led Stackelberg game strategy with dynamic fund outflows is proposed to mitigate such incentive misalignment. Numerical optimization results show that mitigation of the manager-investor conflict of interests is possible with the proposed Stackelberg game strategy.
|Type of resource
|electronic; electronic resource; remote
|1 online resource.
|Zou, Hao, Mr
|Stanford University, Department of Electrical Engineering
|Cioffi, John M
|Cioffi, John M
|Statement of responsibility
|Submitted to the Department of Electrical Engineering.
|Thesis (Ph.D.)--Stanford University, 2011.
- © 2011 by Hao Zou
Also listed in
Loading usage metrics...