<p> An optical time division multiplexing (TDM) scheme using soliton interaction is proposed in the thesis to save the time-bandwidth prduct (TBP). The soliton multiplexer (MUX) consisting of a highly nonlinear fiber (HNLF) combines two adjacent solitons to form a composite soliton, while the soliton demultiplexer (DEMUX) consisting of a similar HNLF restores the component solitons. The case of interaction between identical fundamental solitons is discussed first. However, when this scheme is used in the conventional TDM system, the total bit rate transmitted over the channel is limited by the time interval between the two adjacent component solitons. Therefore, a modified multiplexing scheme using interaction between different solitons is proposed to satisfy more practical engineering applications. The theoretical analysis and numerical simulation results demonstrate that the modified optical TDM scheme offers a higher TBP efficiency and suitable for conventional TDM, which makes it an attractive candidate for meeting the challenge of increasing demand on frequency bandwidth in modern optical communications. </p> / Thesis / Master of Applied Science (MASc)

M.S.

The daily, weekly, and unscheduled tasks for school administrators have increased in number and scope over the years, however surprisingly little is known about what principals do on a day-to-day basis and how this varies across schools. Since the effect of principal leadership behaviors, specifically how principals manage their time to accomplish important tasks, is one key to the success of schools, it is important to understand what effective principals do to accomplish this. The purpose of this study was to find out what the differences are in how principals in high and low-performing middle schools spend their time and to determine what relationships exist between the principal's time-on-tasks and school performance. In the literature review, the researcher identified seven categories of time use to collect and classify time-on-tasks data. The categories include: (a) administration/operations, (b) organization management, (c) day-to-day instruction, (d) instructional program, (e) internal relations, (f) external relations and (g) other (Horng, Klasik, & Loeb, 2010). The researcher collected time-on-tasks data from principals of high and low-performing middle schools in Virginia and analyzed the data to determine what relationships exist between the principal's time-on-tasks and school performance. Data analyses revealed that there are significant differences in the amount of time principals at high-performing schools devote to each of the time-on-tasks categories, as compared to the amount of time allocated by their counterparts at low-performing schools. In this study, principals as a whole and principals in the high-performing subgroup spend the largest percentage of time on tasks related to administration and operations, while principals in the low-performing subgroup spend the largest percentage of time on day-to-day instruction. Data also suggest that time spent on tasks related to internal relations is positively correlated with student performance on mathematics and reading tests. When demographic factors are combined with the time-on-tasks categories, a regression analysis suggests that the strongest contributing factor to mathematics and reading test scores is the socioeconomic status of the school with a strong negative correlation between the percentage of students on free/reduced lunch and test scores for mathematics and reading. / Ed. D.

The fact that traffic congestion upstream of a bottleneck causes a reduction in the discharge flow rate through the bottleneck has been well documented in several empirical studies. However, what has been missing is an understanding of the causes of these empirically observed flow reductions. An identification of these causes is important in order to develop various mitigation schemes through the use of emerging technology. The concept of capacity drop can be introduced at time-independent bottlenecks (e.g. freeways) as well as time-dependent bottlenecks (e.g. signalized intersections). While to the author's knowledge no one has attempted to link these phenomena, the research presented in this thesis serves as a first step in doing so. The research uses the INTEGRATION simulation software, after demonstrating its validity against empirical data, to simulate time-independent and time-dependent bottlenecks in an attempt to characterize and understand the contributing factors to these flow reductions. Initially, the INTEGRATION simulation software is validated by comparing its results to empirically observed traffic stream behavior. This thesis demonstrates that the discharge flow rate is reduced at stationary bottlenecks at the onset of congestion. These reductions at stationary bottlenecks are not recovered as the traffic stream propagates downstream. Furthermore, these reductions are not impacted by the level of vehicle acceleration. Alternatively, the drop in the discharge flow rate caused by time-dependent bottleneck is recoverable and is dependent on the level of acceleration. The difference in behavior is attributed to the fact that in the case of a stationary bottleneck the delay in vehicle headways exceeds the losses caused by vehicle accelerations and thus is not recoverable. In the case of vehicles discharging from a backward recovery wave the dominant factor is the delay caused by vehicle acceleration and this can be recuperated as the traffic stream travels downstream. / Master of Science

Real-time channels are established between a source and a destination to guarantee in-time delivery of real-time messages in multi-hop networks. In this thesis, we propose two schemes to establish real-time channels for soft real-time applications whose timeliness properties are characterized using Jensen's Time Utility Functions (TUFs) that are non-increasing. The two algorithms are (1) Localized Decision for Utility accrual Channel Establishment (LocDUCE) and (2) Global Decision for Utility accrual Channel Establishment (GloDUCE). Since finding a feasible path optimizing multiple constraints is an NP-Complete problem, these schemes heuristically attempt to maximize the system-wide accrued utility. The channel establishment algorithms assume the existence of a utility-aware packet scheduling algorithm at the interfaces. The route selection is based on delay estimation performed at the source, destination, and all routers in the path, from source to destination. We simulate the algorithms, measure and compare their performance with open shortest path first (OSPF). Our simulation experiments show that for most of the cases considered LocDUCE and GloDUCE perform better than OSPF. We also implement the schemes in a proof-of-concept style routing module and measure the performance of the schemes and compare them to OSPF. Our experiments on the implementation follow the same trend as the simulation study and show that LocDUCE and GloDUCE have a distinct advantage over OSPF and accrue higher system-wide utility. These schemes also react better to variation in the loading of the links. Among the two proposed approaches, we observe that GloDUCE performs better than LocDUCE under conditions of increased downstream link loads. / Master of Science

Time-series forecasting has long been studied and remains an important research task. In scenarios where multiple time series need to be forecast, approaches that exploit the mutual impact between time series results in more accurate forecasts. This has been demonstrated in various applications, including demand forecasting and traffic forecasting, among others. Hence, this dissertation focuses on graph-based models, which leverage the internode relations to forecast more efficiently and effectively by associating time series with nodes. This dissertation begins by introducing the notion of graph time-series models in a comprehensive survey of related models. The main contributions of this survey are: (1) A novel categorization is proposed to thoroughly analyze over 20 representative graph time-series models from various perspectives, including temporal components, propagation procedures, and graph construction methods, among others. (2) Similarities and differences among models are discussed to provide a fundamental understanding of decisive factors in graph time-series models. Model challenges and future directions are also discussed. Following the survey, this dissertation develops graph time-series models that utilize complex time-series interactions to yield context-aware, real-time, and probabilistic forecasting. The first method, Context Integrated Graph Neural Network (CIGNN), targets resource forecasting with contextual data. Previous solutions either neglect contextual data or only leverage static features, which fail to exploit contextual information. Its main contributions include: (1) Integrating multiple contextual graphs; and (2) Introducing and incorporating temporal, spatial, relational, and contextual dependencies; The second method, Evolving Super Graph Neural Network (ESGNN), targets large-scale time-series datasets through training on super graphs. Most graph time-series models let each node associate with a time series, potentially resulting in a high time cost. Its main contributions include: (1) Generating multiple super graphs to reflect node dynamics at different periods; and (2) Proposing an efficient super graph construction method based on K-Means and LSH; The third method, Probabilistic Hypergraph Recurrent Neural Network (PHRNN), targets datasets under the assumption that nodes interact in a simultaneous broadcasting manner. Previous hypergraph approaches leverage a static weight hypergraph, which fails to capture the interaction dynamics among nodes. Its main contributions include: (1) Learning a probabilistic hypergraph structure from the time series; and (2) Proposing the use of a KNN hypergraph for hypergraph initialization and regularization. The last method, Graph Deep Factors (GraphDF), aims at efficient and effective probabilistic forecasting. Previous probabilistic approaches neglect the interrelations between time series. Its main contributions include: (1) Proposing a framework that consists of a relational global component and a relational local component; (2) Conducting analysis in terms of accuracy, efficiency, scalability, and simulation with opportunistic scheduling. (3) Designing an algorithm for incremental online learning. / Doctor of Philosophy / Time-series forecasting has long been studied due to its usefulness in numerous applications, including demand forecasting, traffic forecasting, and workload forecasting, among others. In scenarios where multiple time series need to be forecast, approaches that exploit the mutual impact between time series results in more accurate forecasts. Hence, this dissertation focuses on a specific area of deep learning: graph time-series models. These models associate time series with a graph structure for more efficient and effective forecasting. This dissertation introduces the notion of graph time series through a comprehensive survey and analyzes representative graph time-series models to help readers gain a fundamental understanding of graph time series. Following the survey, this dissertation develops graph time-series models that utilize complex time-series interactions to yield context-aware, real-time, and probabilistic forecasting. The first method, Context Integrated Graph Neural Network (CIGNN), incorporates multiple contextual graph time series for resource time-series forecasting. The second method, Evolving Super Graph Neural Network (ESGNN), constructs dynamic super graphs for large-scale time-series forecasting. The third method, Probabilistic Hypergraph Recurrent Neural Network (PHRNN), designs a probabilistic hypergraph model that learns the interactions between nodes as distributions in a hypergraph structure. The last method, Graph Deep Factors (GraphDF), targets probabilistic time-series forecasting with a relational global component and a relational local model. These methods collectively covers various data characteristics and model structures, including graphs, super graph, and hypergraphs; a single graph, dual graphs, and multiple graphs; point forecasting and probabilistic forecasting; offline learning and online learning; and both small and large-scale datasets. This dissertation also highlights the similarities and differences between these methods. In the end, future directions in the area of graph time series are also provided.

Scheduling decisions in soft real-time environments are based on a utility function. The goal of such schedulers is to use a best-effort approach to maximize the utility function and ensure graceful degradation at overloads. Utility Accrual (UA) schedulers use heuristics to maximize the accrued utility. Heuristic-based scheduling do not always yield the optimal schedule even if there exists one because they do not explore the entire search space of task orderings. In distributed systems, local UA schedulers use the same heuristics along with deadline decomposition for task segments. At present, there has been no evaluation and analysis of the degree to which these polynomial-time, heuristic algorithms succeed in maximizing the total utility accrued. We implemented a preemptive, off-line static scheduling algorithm that performs an exhaustive search of all the possible task orderings to yield the optimal schedules. We simulated two important online dynamic UA schedulers, DASA-ND and LBESA for different system loads, task models, utility and load distribution patterns, and compared their performance with their corresponding optimal schedules. Our experimental analysis indicates that for most scenarios, both DASA-ND and LBESA create optimal schedules. When task utilities are equal or form a geometric sequence with an order of magnitude difference in their utility values, UA schedulers show more than 90% probability of being optimal for single-node workloads. Even though deadline decomposition substantially improves the optimality of both DASA-ND and LBESA under different scenarios for distributed workloads, it can adversely affect the scheduling decisions for some task sets we considered. / Master of Science

Yes / We investigated whether changes in low-level image characteristics, in this case spatial frequency, were capable of generating a well-known expansion in the perceived duration of an infrequent “oddball” stimulus relative to a repeatedly-presented “standard” stimulus. Our standard and oddball stimuli were Gabor patches that differed from each other in spatial frequency by two octaves. All stimuli were equated for visibility. Rather than the expected “subjective time expansion” found in previous studies, we obtained an equal and opposite expansion or contraction of perceived time dependent upon the spatial frequency relationship of the standard and oddball stimulus. Subsequent experiments using equi-visible stimuli reveal that mid-range spatial frequencies (ca. 2 c/deg) are consistently perceived as having longer durations than low (0.5 c/deg) or high (8 c/deg) spatial frequencies, despite having the same physical duration. Rather than forming a fixed proportion of baseline duration, this bias is constant in additive terms and implicates systematic variations in visual persistence across spatial frequency. Our results have implications for the widely cited finding that auditory stimuli are judged to be longer in duration than visual stimuli. / Wellcome Trust, UK, the Federation of Ophthalmic and Dispensing Opticians, UK, and the College of Optometrists, UK.

Typescript (photocopy). / Digitized by Kansas Correctional Industries

Call number: LD2668 .T4 1960 C68