SHAREOK Repository
Not a member yet
49261 research outputs found
Sort by
Compression techniques for extreme-scale graphs and matrices: sequential and parallel algorithms
Get PDFA graph G = (V, E) is an ordered tuple where V is a non-empty set of elements called vertices (nodes), and E is a set of an unordered pair of elements called links (edges), and a time-evolving graph is a change in the states of the edges over time. With the growing popularity of social networks and the massive influx of users, it is becoming a challenging task to store the network/graph and process them as fast as possible before the property of the graph changes with the graph evolution.
Graphs or networks are a collection of entities (individuals in a social network) and their relationships (friends, followers); ways to represent a graph can help how the information could be extracted. The increase in the number of users increases the relationship the user has, which makes the graphs massive and nearly impossible to store them in friendly structures such as a matrix or an adjacency list. Therefore, an exciting area of research is storing these massive graphs with a smaller memory footprint and processing with very little extra memory.
But there is always a trade-off with time and space; to get a small memory footprint, one has to remove the redundancy rigorously, which consumes time. In the same way, when traversing these tight spaces, the time required to query also increases compared to a matrix or an adjacency list.
In this dissertation, we provide the encoding technique to store the arrays in the Compressed Sparse Row (CSR) data structure and extend the encoding to store time-evolving graphs in the form of a CSR. We also propose combinations of two structures (CSR + CBT) to store the time-evolving graphs and to improve the time and space trade-off. Encoding also enables one to access a node without decompressing the entire structure, which means that the data structure can be accessed.
We then provide four ways to store multi-dimensional data, which represents intricate relations within the social network. Once the data are stored in compressed format, it is important to provide algorithms that support the structures. One such computation which is the basis for any graph algorithm is matrix-multiplication. We now extend our work to perform value-based matrix multiplication on compressed structures. We test our algorithm on extremely large matrices, in the order of 100s of millions with various levels of sparsity. Using matrix-matrix multiplication and keeping the theme of storing the data in small spaces, we propose another way of compression is through the dimensionality reduction, which is referred to as Matrix Factorization.
Performing any of these operations on a compressed structure without decompressing would be time consuming. Therefore, in this dissertation, we introduce a parallel technique to construct the graph and also run a list of queries using the querying algorithms, such as fetching neighbor or edge existence in parallel. We also extend our work to propose parallel time-evolving differential compression of CSR using the prefix sum approach
Applying Dynamic Mode Decomposition to Interconnected Systems for Forecasting and System Identification
Get PDFDynamic Mode Decomposition (DMD) describes a family of dynamical systems analysis approaches that approximate complex, likely non-linear behaviors with a low-rank linear operator. DMD has traditionally been used in a systems-identification context and was originally developed as a method of modeling fluid flows using the Koopman Operator. In contrast to these original applications, this work explores DMD's ability to produce high-fidelity forecasts using small training sets in an effort to flexibly model two complex, real-world systems. In particular, a novel, iterative implementation of DMD is tested and validated on 18 years of trading price data for constituent companies of the S\&P 500 and on 2 years of per-capita COVID-19 case counts throughout the continental US. The novel combination of DMD with blocked time-series cross-validation described in this work was found to consistently produce forecasts with an average MAPE of approximately 0.1 (in the case of the financial model) and RMSE of approximately 0.2 cases per 1000 citizens (for the COVID-19 model). In addition to reliably predicting the complex behaviors characteristic of real-world systems, this approach was leveraged to identify robust, distinct dynamical trends and construct networks which provided insights into central system elements. This work has illustrated the utility of applying DMD in an iterative approach facilitates forecasting accuracy across a variety of systems without compromising its ability to uncover fundamental characteristics of these underlying systems
Microwave filters for next generation radio frequency transceivers
Get PDFIncreased data rates in wireless communications enforce unprecedented performance metrics on the front-end filters to operate in crowded spectral bands. These requirements include strong selectivity, low insertion loss, and good out-of-band (OOB) rejection in addition to the applicability in complementary metal oxide semiconductor (CMOS) integrated circuit layouts. The acoustic wave (AW) resonator based filter design technology has gained a very important role in the on-chip filter design techniques due to chip-scale physical resonator sizes and the ability of achieving high quality factor values at microwave frequencies. However, conventional synthesis methods used in the design of AW resonator based microwave filters suffer from limited achievable fractional bandwidth (FBW) and weak OOB rejection. The origin of these issues is the limitations on increasing the electromechanical coupling coefficient (kt2) of the resonators, which is an intrinsic property of the piezoelectric material in its design. This dissertation proposes a new class of hybrid acoustic-electromagnetic (Hybrid-ACEM) filters to overcome both of the aforementioned limitations of AW resonator-based filters. In other words, the main goal of this new topology is to maximize the ratio between the achievable FBW and the required kt2. This is achieved by employing one or two electromagnetic (EM) resonators that are placed at purposefully selected stages within the design. In addition, cross-coupling mechanisms are systematically used to reduce the required electromechanical coupling coefficient in certain filter orders. Altogether, the proposed method can achieve much larger FBW values and stronger OOB rejection compared to the conventionally synthesized ladder acoustic wave filters. The effect of finite quality factor of the EM resonators is analyzed. A new algorithm to convert extracted-pole sections to Butterworth-Van-Dyke (BVD) model for large FBW values is also presented. It has been shown in the simulations that FBW-to-kt2 ratios of four or above is achievable with this method. As a proof-of-concept, a sixth-order hybrid canonical prototype with a center frequency of 2.67 GHz and 11.2% FBW is designed and fabricated. The acoustic wave resonators used in the fabrication have kt2 values of 3.5%. The fabricated prototype proves the validity of the proposed method for achieving FBW values of 30% with required kt2 values of 7.5%, which is available with the common aluminum nitride (AlN) based bulk acoustic wave resonator technologies of today. The developed technique opens a new pathway to reduce the limitations of integrating microwave filters for future fully on-chip microwave transceivers
Designing data-aided demand-driven user-centric architecture for 6G and beyond networks
Get PDFDespite advancements in capacity-enhancing technologies like massive MIMO (multiple input, multiple output) and intelligent reflective surfaces, network densification remains crucial for significant capacity gains in future networks such as 6G. However, network densification increases interference and power consumption. Traditional cellular architectures struggle to minimize these without compromising service quality or capacity, which necessitates a shift to a user-centric radio access network (UC-RAN).
The UC-RAN approach offers additional degrees of freedom to ease the spectral-energy efficiency interlock while improving the service quality. However, its increased degrees of freedom make its optimal design and operation more challenging. This dissertation introduces four novel approaches for UC-RAN optimal design and operation. The objectives include mitigating interference, reducing power consumption, ensuring diverse user/vertical service quality, facilitating proactive network operation, risk-aware optimization, adopting an open radio access network, and enabling universal coverage.
First, we construct an analytical framework to assess the effects of incorporating Coordinated Multipoint (CoMP) technology into UC-RAN to reduce interference and power consumption. We use stochastic geometry tools to derive expressions for network-wide coverage, spectral efficiency, and energy efficiency as a function of UC-RAN Configuration and Optimization Parameters (COPs), including data base station densities and user-centric service zone sizes.
While the analytical framework provides insightful performance analysis that can guide overall system design, it cannot fully capture the dynamics of a UC-RAN system to enable optimal operation. Next, we present a Deep Reinforcement Learning (DRL) based method to dynamically orchestrate the UC-RAN service zone size to satisfy varying application demands of various service verticals during its operation. We define a novel multi-objective optimization problem that fairly optimizes otherwise conflicting key performance indicators (KPIs).
DRL's practical adaptation by the industry remains thwarted by the risk it poses to the safe operation of a live network. To address this challenge, we propose a digital twin-enabled approach to enrich the DRL-based optimization framework, ensuring risk-aware COP optimization. We use Open Radio Access Network standards-based simulations to show that the proposed risk-aware DRL framework can maximize system-level KPIs while maintaining safe operational requirements.
Lastly, we propose a hybrid model of aerial and terrestrial UC-RAN deployment to ensure universal coverage. We assess the impact of aerial base station parameters on system-level KPIs, providing a quantitative analysis of the advantages of a hybrid over a solely terrestrial UC-RAN. We develop a robust multi-objective function solvable via our DRL-based framework to balance and optimize these KPIs in a hybrid UC-RAN.
Our extensive analytical and system-level simulation results suggest that these contributions can foster the much-needed paradigm shift towards demand-driven, elastic, and user-centric architecture in emerging and future cellular networks
Analysis and Control of Stick-Slip using a Scaled Experimental Drilling Rig
No full textDrilling is one of the most challenging and expensive processes involved in oil and gas production. Due to the destructive nature of the drilling process, dysfunctions are common during drilling. These dysfunctions can increase the non-productive time (NPT) greatly, cause loss of revenue and pose a safety concern. Avoiding and controlling these dysfunctions is essential for safe and efficient drilling.
Friction from the wellbore and the rock formation induces vibrations in the drillstring. The drillstring vibrations can be classified into three modes: axial, lateral and torsional. Stick-slip vibrations are a type of torsional vibrations which occur very frequently during drilling. These vibrations are responsible for decreasing the rate of penetration (ROP), causing damage to the drilling equipment and greatly reducing the life of a drill bit. Different control algorithms have been proposed in the last few decades but only one of them (Shell Soft Torque) has seen industry-wide adoption. Now, with the increase in the complexity of well design and increasing depths, the limitations of Soft Torque have been highlighted. Therefore, extensive research is still underway to design a new stick-slip con-trol/mitigation approach.
The main aim of this study was to analyze and control stick-slip vibrations in a drillstring using a downscaled experimental drilling rig setup. To achieve this, an experimental setup was designed and constructed. Thereafter, strategies to control/mitigate stick-slip were devised and tested on the constructed setup.
The results showed that surface parameters alone may not be sufficient to design a robust stick-slip mitigation system. Access to downhole data, even at low frequency, can significantly help in the development of a stick-slip mitigation system
Framework Literacy: Understanding, Implementing and Integrating the ACRL Framework
No full textThe ACRL Framework for Information Literacy was introduced in 2015, and since then library professionals have developed a myriad of ways of incorporating the Framework, including active methods of teaching these concepts directly to learners, or working these concepts into previously developed practices. How can we integrate the Framework into library instruction, reference and other services? Over the last eight years, how have Oklahoma librarians already done so? Where have we found success? Is there anything missing in the Framework? Is it time for the ACRL to update it? Join in on an exploration of the concepts within the Framework, where learners will evaluate the language of the Six Frames, and discover methods to apply these ideas into library services.N
Inflation in Iran: An Empirical Assessment of the Key Determinants
Get PDFPurpose: To study the key determinants of chronically high inflation in Iran.Design/Methodology/Approach: Relying on annual data from 1978 to 2019, we employ an Auto-Regressive Distributed Lag Model (ARDL) and Error Correction Model (ECM) to study the inflationary effects of monetary and fiscal policies as well as exchange rate swings and sanctions intensification.Findings: We find that increase in money supply, depreciation of nominal exchange rate, increase in fiscal deficit, and intensification of sanctions are among the key drivers of inflation in Iran. Their impact is profound in the long run, but in the short run only money supply and currency depreciation are significant. Also, when exploring the inflation in different components of Consumer Price Index (CPI), we find robust long- and short-run effects from money supply and exchange rate, while the effects of fiscal deficit and sanctions vary across different components.Originality/Value: We contribute to the literature by setting apart the long- vs. short-run effects of key variables on inflation in Iran. We also employ improved measures of fiscal deficit and sanctions that are shown to be of significance in the long run. Lastly, we go beyond the aggregate index and examine the variations in different CPI components.Yes© Journal of Economic Studies, Emerald Publishing Limited. DOI: 10.1108/JES-07-2022-0370. This author accepted manuscript (AAM) is provided for your own personal use only. It may not be used for resale, reprinting, systematic distribution, emailing, or for any other commercial purpose without the permission of the publisher
Hope-Centered Leadership
Get PDFHope has been found to be a strong predictor of employee well-being. However, until now, no scale existed to measure whether leaders were activating and operationalizing hope within employees. The Hope-Centered Leadership (HCL) scale addresses that void. Using the framework of Snyder’s established Hope Theory and Yukl’s Taxonomy of Leadership Behaviors, this study constructed a conceptualization of HCL, developed and validated the HCL measure, and demonstrated HCL as a resource in the Job Demands-Resources model. First, HCL was defined as behaviors that activate and nurture hope through setting task-oriented goals, navigating change-oriented pathways, and cultivating relations-oriented agency. Second, a sample of 340 teachers was utilized to validate the measure. Confirmatory factor analysis confirmed an a priori first-order structure consisting of nine items. Convergent and discriminate validity was confirmed through structural equation modeling (SEM) utilizing hope, collective hope, and trust. Third, HCL was tested using a sample of 501 individuals within the Job Demands-Resources model to determine if HCL served as a job resource. SEM results indicated that HCL did serve as a job resource through positive correlations with collective hope and workplace well-being while reducing the effects of job demands (i.e., exhaustion, abusive supervision) leading to burnout. Consequently, practitioners now have a valid measure built on established theory for use in determining if leaders are fostering hope within employees through goals, pathways, and agency
Study of hot carrier dynamics and relaxation in metal halide perovskites
Get PDFMetal halide perovskites are a promising class of solution-based semiconductors
with significant potential for optoelectronic devices, including hot carrier solar cells
that are proposed to achieve higher efficiency than conventional solar cells. These
materials possess high light absorption coefficients, long carrier diffusion lengths,
and high defect tolerance, which leads to high power conversion efficiencies in
solar cells. However, the intrinsic material instability of perovskite-based solar
cells limits their performance. Understanding the carrier dynamics and stability
of the material is critical to the development of solar cell physics, particularly
in the case of hot carrier solar cells, as excited carriers with energy greater than
the bandgap energy can potentially generate higher voltage and current than
traditional solar cells. Unfortunately, these hot carriers also tend to rapidly lose
their energy through carrier-phonon scattering, resulting in poor energy conversion.
Perovskites have shown significant potential for hot carrier solar cells in recent
years due to their slow carrier cooling rate. However, a better understanding of
carrier dynamics is critical for developing efficient perovskite-based solar cells.
The effect of changing the different components of perovskite compounds was
studied in a series of metal halide perovskites with varying optical properties
using temperature-dependent photoluminescence, power-dependent photoluminescence,
and ultrafast transient absorption techniques to investigate the carrier
dynamics. The results suggest that the slow cooling of carriers in metal-halide perovskites results from the intrinsic low thermal conductivity of all metal-halide
perovskites. This finding indicates that the phonon energy, exciton binding energy,
and interaction strength have little effect on the cooling of carriers. This study
provides valuable insights into the fundamental understanding of carrier dynamics
in perovskite-based solar cells, which can guide the development of more efficient
and stable devices.
In addition we also studied the presence of hot carriers in a stable metal
halide perovskite system under steady-state conditions. The results showed clear
evidence of hot carriers in the device, but their behavior was strongly dependent
on temperature and competition with photo-induced halide segregation. These
findings provide valuable insights into the behavior of hot carriers in metal halide
perovskite devices for the development of more efficient and stable hot carrier
solar cells.
Finally, our study on the 2D Ruddlesden-Popper perovskite (EPEA)2PbI4 using
temperature and power dependent photoluminescence and transient absorption
spectroscopy showed the presence of multiple excitonic complexes and carrier
redistribution mediated by power and/or temperature. Moreover, we observed
extremely long-lived dark states in transient absorption, which play a significant
role in the photoluminescence and absorption dynamics of (EPEA)2PbI4. These
findings contribute to our understanding of perovskite material behavior and
could aid in the development of more efficient optoelectronic devices
Losing My Political Motivation: Trends in Voting and Protesting Expectations of U.S. High School Seniors
Get PDFThe purpose of this descriptive study is to analyze trends in the political engagement plans of high school seniors from 1976 to 2019. Using the Monitoring the Future dataset, I address the following research questions: Have expectations of voting and protesting changed for U.S. high school seniors over time? If so, do patterns of change vary by gender and racial-ethnic identity? Findings show that voting and protesting expectations have declined for young women and men during the time period. Young women are consistently more likely to expect to vote than young men throughout the period, but it is less significant of a difference over time, suggesting the gap may be closing. While voting expectations have declined over time for all three of the racial-ethnic identity groups in my study, white individuals are more likely to expect to vote over time than Black and Hispanic individuals. A gender gap in protesting that favored young men has declined over time. Results also indicate that expectations of protesting have decreased over time (2005-2019) for young white and Hispanic individuals but not for Black individuals