MS, University of Cincinnati, 2016, Engineering and Applied Science: Computer Science

Today, large number of companies are migrating to the cloud, leaving behind the concept of maintaining traditional data centers and servers. The main reasons for this migration include reduced capital costs, reduced expenditure on infrastructure and ease of accessibility. With the increasing demand for Cloud Computing and the changing needs of users, a need to make the services on the cloud dynamic in nature is essential. However, dynamic services require constant costly updates and highly meticulous configurations. One such dynamic service offered by Amazon Web Services (AWS) is Auto Scaling Groups (ASGs). With this service, AWS facilitates automatic scale up and scale down on the count of servers (instance resources) based on the ASG policies and conditions set by the users. A small misconfiguration or a build failure associated with the Amazon Machine Image (AMI) could cause the dynamism to occur when not actually needed. Since users are charged for the instances by the hour, unnecessary costs occur even if the usage is for as less as a minute. This situation of unnecessary launch and termination of instances is termed as “flaps” and can be compared to oscillations in signals. To prevent energy dissipation in case of oscillating signals, damping of signals is performed. This is similar to the problem of flapping in ASGs. We have come up with a software called AWS Flap Detector as a solution to this problem. AWS Flap Detector efficiently detects and reports flapping Auto Scaling Groups and paves the way for correction. This in turn helps prevent unnecessary resource allocation and billing.

Doctor of Philosophy, The Ohio State University, 2021, Computer Science and Engineering

Sustainable computing, dark silicon and approximate computing have ushered a new era in which some processing capacity is available only as ephemeral bursts, a technique called computational sprinting. Computational sprinting speeds up query execution by increasing power usage, dropping tasks, precision scaling, and etc. for short bursts. Sprinting policy decides when and how long to sprint. Poor policies inflate response time significantly. However, sprinting alters query executions at runtime, creating a complex dependency between queuing and processing time. Sprinting can speed up query processing and reduce queuing delay, but it is challenging to set efficient policies. As sprinting mechanisms proliferate, system managers will need tools to set policies so that response time goals are met. I provide a method to measure the efficiency of sprinting policies and a framework to create response time models for sprinting mechanisms such as DVFS, CPU throttling, cache allocation, and core scaling. I compared sprinting policies used in competitive solutions with policies found using our models.

Doctor of Philosophy (PhD), Wright State University, 2018, Biomedical Sciences PhD

Synaptic plasticity, the ability of neurons to modulate their inputs in response to changing stimuli, occurs in two forms which have opposing effects on synaptic physiology. Hebbian plasticity induces rapid, persistent changes at individual synapses in a positive feedback manner. Homeostatic plasticity is a negative feedback effect that responds to chronic changes in network activity by inducing opposing, network-wide changes in synaptic strength and restoring activity to its original level. The changes in synaptic strength can be measured as changes in the amplitudes of miniature post-synaptic excitatory currents (mEPSCs). Together, the two forms of plasticity underpin nervous system functions such as movement, learning and memory, and perception, while preventing pathological states of hyper- or hypoactivity that could occur if network activity were not maintained. The current hypothesis of homeostatic plasticity states that mEPSC amplitudes exhibit uniform multiplicative scaling, a transformation in which the amplitudes are scaled up or down globally by a multiplicative factor. This hypothesis constrains the possible mechanism of homeostatic plasticity, which remains unknown despite intensive study. Here, we compare an experimental data set previously collected in our laboratory to the results of an empirical simulation of uniform multiplicative scaling and conclude that the homeostatic increase in mEPSC amplitudes in our data is not uniform. We develop and validate a novel method, comparative standardization, for calculating the scaling transformation between treated and untreated mEPSC amplitudes and identifying the transformation as either uniform, divergent, or convergent. When applied to our experimental data, comparative standardization finds divergent scaling, in which the homeostatic effect increases with synaptic strength, causing the control and treated mEPSC amplitude distributions to diverge. The divergent scaling transformation computed by compara (open full item for complete abstract)

Master of Science (MS), Ohio University, 2017, Electrical Engineering & Computer Science (Engineering and Technology)

As the relentless quest for higher throughput and lower energy cost continues in heterogenous multicores, there is a strong demand for energy-efficient and high-performance Network-on-Chip (NoC) architectures. Heterogenous architectures that can simultaneously utilize both the serialized nature of the CPU as well as the thread level parallelism of the GPU are gaining traction in the industry. A critical issue with heterogenous architectures is finding an optimal way to utilize the shared resources such as the last level cache (LLC) and NoC without hindering the performance of either the CPU or the GPU core. Photonic interconnects are a disruptive technology solution that have the potential to increase the bandwidth, reduce latency, and improve energy-efficiency over traditional metallic interconnects. In this thesis, we propose a CPU-GPU heterogenous architecture called SHARP (Shared Heterogenous Architecture with Reconfigurable Photonic Network-on-Chip) that combines CPU and GPU cores around the same router. SHARP architecture is designed as a Single-Writer Multiple-Reader (SWMR) crossbar with reservation-assist to connect CPU/GPU cores. The architecture consists of 32 CPU cores and 64 GPU computational units. As network traffic exhibits temporal and spatial fluctuations due to application behavior, SHARP can dynamically reallocate bandwidth and thereby adapt to application demands. In this thesis, we propose to dynamically reallocate bandwidth and reduce power consumption by evaluating buffer utilization. While buffer utilization is a reactive technique that deals with fluctuations in application demands, we also propose a proactive technique wherein we use machine learning (ML) to optimize the bandwidth and power consumption. In ML, instead of predicting the buffer utilization, we predict the number of packets that will be generated by the heterogenous cluster. Simulation results where evaluated using PARSEC 2.1 and SPLASH2 benchmark suits for the CPU and Ope (open full item for complete abstract)

PHD, Kent State University, 2009, College of Arts and Sciences / Department of Mathematical Sciences

There is an extensive literature on linear maps and operators on Banach spaces. However, a corresponding non-linear theory is still in its beginning. We study non-linear maps between subsets of the classical Banach spaces and give estimates for the Banach-Mazur Lipschitz norm for some of the natural non-linear maps between unit spheres of these spaces. We show that the Banach-Mazur norm of these maps is much larger than the corresponding linear norms. We also study projections onto unit spheres of Banach spaces, we show that the scaling down projection is the best projection in L1(0,1), and we give estimates for the Lp-spaces, p>1.

Master of Science (M.S.), University of Dayton, 2024, Electrical Engineering

RGB imagery provides detail which is usually sufficient to perform computer vision tasks. However, images taken in low-light appear vastly different from well-lit imagery due to the diversity in light intensity. Polarimetric data provides additional detail which focuses on the orientation of the light rather than intensity. Scaling our classic RGB images using polarimetric data can maintain the RGB image type, while also enhancing image contrast. This allows transfer learning using pre-trained RGB models to appear more feasible. Our work focuses on developing a large dataset of paired polarimetric RGB images in a highly controlled laboratory environment. Then, we perform transfer learning on a pre-trained image segmentation model with each of our image product types. Finally, we compare these results in both well-lit and low-light scenarios to see how our polarimetrically enhanced RGB images stack up against regular RGB images.

Master of Computing and Information Systems, Youngstown State University, 2024, Department of Computer Science and Information Systems

Graph Neural Networks (GNNs) are increasingly used for challenging tasks such as particle track reconstruction and high-complex graph-based modeling. The training of such large-scale models involves significant challenges in terms of scalability and efficiency, especially when dealing with vast datasets, which are common in high-energy physics. This study is based on scaling GNNs that have shown significant potential in reconstructing particle tracks in densely packed environments like those expected at the HL-LHC. We have utilized Distributed Data Parallelism (DDP) and conducted scaling experiments with the TrackML benchmark dataset to analyze the impact of the quantity of GPUs. Our findings reveal improvements in scalability, reductions in training times, and memory requirements without significantly affecting model efficiency. Additionally, we analyzed the effects of increased model capacity and training dataset size on a single GPU. Expanding hidden layer sizes and dataset size led to better generalization, enhanced efficiency, and improved performance metrics, though at the cost of higher memory usage. These findings highlight the potential of scalable GNN methods for real-time applications when dealing with data-intensive environments.

Master of Arts, The Ohio State University, 2007, Graduate School

Master of Science, The Ohio State University, 2008, Graduate School

Doctor of Philosophy (PhD), Ohio University, 2024, Civil Engineering (Engineering and Technology)

Colloidal aggregation is a critical phenomenon influencing various environmental processes. However, limited research has been conducted on the aggregation of particles with heterogeneous physical and chemical properties, which are more representative of practical environmental systems than homogeneous particles. The central hypothesis of this dissertation is that primary particle size polydispersity along with chemical and material heterogeneity of primary particles exert non-trivial effects on the aggregate growth rate and the fractal dimensions of aggregates. In this dissertation, the aggregation and stability of heterogeneous nano- and sub-micrometer particles in aqueous systems were investigated using in situ microscopy and image analysis. Initially, the study examined the growth kinetics and structures of aggregates formed by polystyrene microplastics in mono- and bidisperse systems. Findings indicated that while the primary particle size distribution did not affect the scaling behavior of aggregate growth, it delayed the onset of rapid aggregation. Structural analysis revealed a power law dependence of the aggregate fractal dimension in both mono- and bidisperse systems, with mean fractal dimensions consistent with aggregates from diffusion-limited cluster aggregation. The results also suggested that aggregate fractal dimension was insensitive to shape anisotropy. The dissertation further explored the structure of DLCA aggregates in heterogeneous systems composed of particles with varying sizes, surface charges, and material compositions. The fractal dimensions of DLCA aggregates in these heterogeneous particle systems were similar, ranging from 1.6 to 1.7, and consistent with theoretical predictions and experimental evidence for homogeneous DLCA aggregates. This confirmed the universality of aggregate structures in the DLCA regime, regardless of particle composition. Additionally, a scaling relationship was demonstrated between aggregat (open full item for complete abstract)

PhD, University of Cincinnati, 2024, Education, Criminal Justice, and Human Services: School Psychology

Early childhood education (ECE) teachers who are new to the profession face challenges related to a lack of training and preparation, sustainable support within settings, and professional learning opportunities. Structured peer support can decrease the burden on early career ECE teachers while increasing aspects of well-being. One such structured peer support is structured peer supervision and consultation. While previous research has evaluated the effects of structured peer supervision and consultation on applied communication skills, aspects of well-being, and structures of consultation, the participants have been primarily been school psychology trainees. Therefore, the purpose of the current study was to evaluate the effectiveness of a Preschool Peer Consultation (PPC) intervention package on applied communication skills, well-being, and structures of peer consultation for early career ECE teachers. The study used a single-case withdrawal design to evaluate the PPC intervention package. Results of the study provide preliminary support that the PPC intervention package resulted in an overall increase in teachers' perception of well-being as well as an increase in latency to one communication skill, offering advice.

Master of Sciences (Engineering), Case Western Reserve University, 2024, EMC - Aerospace Engineering

In icing conditions, ice accretes on aircraft components which impacts performance and stability. Researchers simulate icing in wind tunnels to understand icing physics and to develop methods of mitigating its effects. This research investigates the capabilities of the Adaptive Icing Tunnel (AIT), a small-scale tunnel designed to conduct economical and quicker tests. A test plan is developed to characterize the temperature in the AIT including objectives, instrumentation, methodology, and test matrices. Four drop sizing methods are developed that use scaling physics, LEWICE, a cloud droplet probe (CDP), and a rotating multi-cylinder. For specified ice shapes, the scaling and LEWICE methods result in average differences between the calculated and actual MVDs of 24.5% and 71.9% respectively. Methodologies and test plans are developed to use the CDP and the rotating multi-cylinder. The test plans designed will be used during the initial aerothermal and icing cloud characterizations of the AIT.

Master of Sciences, Case Western Reserve University, 2023, EMC - Mechanical Engineering

This thesis describes the design of a set of robotic rat hindlimbs scaled up to 2.5 times the size of the rat. The design is inspired by a previous model from within our lab, but includes a variety of improvements to further the utility and biological accuracy of the model. The robot is comprised of two legs with four motors each to actuate sagittal rotations of the hip, knee, and ankle joints as well as an internal hip rotation. The motor's torque, inertial, viscous, and stiffness properties are characterized for dynamic scaling to be properly implemented in the future control scheme. With direct position commands, the robot's joint movements are able to reflect those of the rat, proving its validity as a test bed for the implementation of future neural control schemes. A previously developed synthetic nervous system control scheme was adapted for and tested on a simulation of the rat hindlimb at animal scale, 2.5 times scale, and 2.5 times scale with joint motor actuation prior to being used to control the robot rat hindlimbs. Concurrent neural simulation and motor control were achieved.

MDES, University of Cincinnati, 2023, Design, Architecture, Art and Planning: Design

The study evaluates the service ecosystem related to scaling a food and beverage consumer packaged goods (CPG) business. It employs service design to de-risk a nascent entrepreneur's pathway from mental concept to product distribution at a national retailer. A service design workshop was developed using semi-structured interviews and ecosystem mapping as data collection tools. This product delivers a personalized visual compass that entrepreneurs can reference. It informs strategic decision-making relevant to the process of scaling a CPG business.

PHD, Kent State University, 2023, College of Arts and Sciences / School of Biomedical Sciences

External jaw morphology has been linked to diet and feeding behavior in primates, where biomechanical significance of condylar trabeculae is still being evaluated. Thus, the structural variation of condylar trabeculae remains largely unknown. I investigate trabecular architecture of the mandibular condyle to assess size-related variation, relationship to feeding behavior, and deterioration secondary to edentulism. I used µCT images of mandibular condyles of adult strepsirrhine primates, adult and neonatal callitrichid primates, and of elderly edentulous and dentate humans. For the non-human primates, the entire trabecular volume was processed using Avizo 8.0 to remove cortical bone. For the human samples, cubed sections were used and compared to cubed sections from the twelfth thoracic vertebra and the distal radius. VOIs were imported into BoneJ for measurement of trabecular parameters. When appropriate, phylogenetic comparative methods were used to address issues of phylogenetic non-independence. Similar to nonhuman primate postcranial bones, the mandibular condyle of strepsirrhine primates exhibits negatively allometric scaling with no alterations in the two features which should most affect load resistance: bone volume fraction and degree of anisotropy. This suggests primate trabecular architecture appears to have generalized scaling trends and trabeculae do not appear to undergo size-related increases in load resistance abilities. This study is the first evidence that the trabeculae of the mandibular condyle scale similarly to those of the postcranium. I compared ontogeny of condylar trabeculae of the tree-gouging common marmoset (Callithrix jacchus) and the non-gouging cotton-top tamarin (Saguineus oedipus). Several growth trends suggest the marmoset does not have a superior load resistance ability than its nongouging relative: bone volume fraction decreased in the marmoset while increasing in the tamarin; spacing increased and connectivity decreased in (open full item for complete abstract)

Master of Science, The Ohio State University, 2022, Environment and Natural Resources

The population of the Central Ohio region is increasing largely due to better economic prospects. The need for housing and related developments will likely go up as the population grows. Most of Ohio's forests are privately owned, and the anticipated developments could impact the current environment by altering the land use of privately owned woodlands. Landowner-level factors impacting changes in land cover and use are largely neglected while predicting these trends. In the first study, private woodland owners were surveyed in multiple counties in Central Ohio on their ownership characteristics, motivations for owning woodlands, demographic factors, and familiarity with ecosystem services account for those factors. The data was analyzed using a binary logistic regression model to identify key elements influencing woodland owners' willingness to sell their property at various price points. The study found that the choice to sell a property was significantly influenced by the landowners' age and residency on the property. Private woodland owners who owned their properties for hunting and amenity values were more likely to sell them. Additionally, landowners aware of the forest's capacity to clean the air expressed less interest in selling their property. On the other hand, landowners who used woodland for recreational activities were less likely to sell. The second study surveyed private woodland owners to determine their preferences for a hypothetical conservation program utilizing binary choice experiments and best-worst choice profiles. Woodland owners were asked to select the best and worst attributes of different programs and their willingness to enroll. Best-Worst scores, Conditional logistic, and Random Effects logistic regression were used to explain woodland owners' priorities. Best-Worst scores show that the highest revenue ($100 acre/year) was the most selected attribute in all choice profiles. A non-profit program structure and no withdrawal penalty are m (open full item for complete abstract)

Doctor of Philosophy, University of Akron, 2022, Mechanical Engineering

Natural convection in laterally heated vertical cylindrical enclosures (LHVCE) has been studied in the past; however, few studies have thoroughly investigated the characteristic length scales. The studied parameters of this enclosure included: the heated and cooled length, diameter, aspect ratio, hot cold wall temperature difference, and the fluid properties. The characteristic length and form of two correction functions were derived from a logical set of assumptions based upon the enclosure configuration. Utilizing the derived dimensionless functions and length scale in conjunction with numerical simulation results, a best fit correlation was formed. The correlation, a first of its kind for this geometry, successfully predicted heat transfer and the flow regime changes from laminar to turbulence. The correlation and the underlying foundation from which it was formed showed to be in agreement with other similar studies. This study then proceeded from enclosures to internal reactor configurations containing baffles. The baffles were grouped into three geometries: rings, single hole openings, and a uniquely shaped (flow enhancing) baffle. The dimensions of each baffle were parametrically varied, typically by scaling the openings; and the velocity and temperature contour results of the simulations are presented. It was found that the central hole and shaped baffle exhibited more desirable flow patterns and thermal environments than the ring baffles. The final portion of the study investigated the effect of the porous media within the enclosure. The porous media cannot be studied independently, thus, two baffles from the previous investigation were utilized for the porous media. Three porous media configurations were studied: homogenous block, extruded concentric rings, and disks. Each of the basic configurations were laid out with a set of scaling parameters and were simulated parametrically in conjunction with the two baffles. The resistance of the porous media was (open full item for complete abstract)

MS, University of Cincinnati, 2021, Engineering and Applied Science: Computer Science

Graph Embedding alternatively known as Representation learning on Graphs has gained a lot of significance in many machine learning applications such as Classification, Prediction, and recommender systems. Many recent methods have developed ways to learn the features and structure of a graph in low dimensional space, however, most of them only focus on the topological information. The extensive node or edge content information available in structured tabular data is only used by data science algorithms separately. In our work, we aim to combine the information available in the form of rich textual or other demographic node attribute information to add context to the graph entities and their interactions. We demonstrate that this refinement of the node embeddings to use the information available in multiple contexts enhance the feature learning in graphs to perform better in Visualization as well as Model building in an unsupervised, generalized way. Specifically, our framework 1. Aims to Combine the node embeddings obtained from the flexible Random walk technique that learns low-dimensional features from network data as well as the node attribute data. 2. Provides a novel and generic approach to refine the Node embeddings by using the Multi-Dimensional Scaling Technique. 3. Compares the visualizations on the refined embeddings in 2D space generated by UMAP and traditional Multi-Dimensional scaling. 4. Provides a way to overcome the transductive nature of Multi Dimension scaling techniques by predicting the refined embeddings for the Unseen/new data using the Triangulation method. We have conducted experiments on 3 real-world datasets and evaluated the efficacy of the final embeddings in low dimensional data separability as well as in multi-label classification. We achieved a maximum classification model accuracy improvement of 441.3% or a minimum of 2.9% when we use combined embeddings overall. We applied our generic framework (open full item for complete abstract)

Master of Science in Engineering, University of Akron, 2021, Engineering

This research quantifies the high-pressure rheological performance of various jet engine turbine oils and helicopter transmission oils. The jet engine oils are classified against both the MIL-PRF-23699 HTS and AS 5780 HPC lubricant performance specifications. The helicopter transmission oils are classified against the DOD-PRF-85734 lubricant performance specification. Rheological properties include the low shear viscosity as a function of temperature and pressure, apparent shear viscosity as a function of temperature, pressure, and shear stress, and lubricant relative volume as a function of temperature and pressure. The low shear viscosity was obtained with a set of pressurized falling body viscometers capable of measuring viscosity at pressures on the order of 1GPa. Apparent shear viscosity was observed using a high pressure Couette viscometer capable of subjecting fluids to 20MPa shear stress. Lastly, density was determined using a relative volume bellows which enables density measurements up to 350MPa of pressure. In addition to the rheological properties, the lubricant's maximum traction coefficients were investigated using a full film traction test. Models were regressed to describe each lubricant's rheological properties. These include the Williams, Landel, and Ferry (WLF) Modified Yasutomi viscosity model, the Modified Carreau Yasuda and Double Modified Carreau Yasuda shear viscosity models, as well as both the Tait and Murnaghan relative volume models used in describing the lubricants' compressibility. In addition to the stated models, thermal aspects of the lubricants were deduced based on scaling rules available in literature. By quantifying the rheological properties of these lubricants, in addition to their thermal characteristics, bearing and gear suppliers can utilize the advanced models presented herein to tailor internal geometries that exploit the lubricants' response to pressure, temperature, and shear stress which In turn, further reduces frictio (open full item for complete abstract)

Master of Science (MS), Wright State University, 2021, Earth and Environmental Sciences

The cumulative number versus floe area distribution of seasonal ice floes from four satellite images covering the Summer season (November - February) in the Weddell Sea Antarctica during the summer breakup and melting is fit by two scale-invariant power scaling regimes for the floe areas ranging from 7 to 20 x 108 m2. Scaling exponents, β, for larger floe areas range from -1.5 to -1.7 with an average of -1.6 for floe areas ranging from 6 x 106 to 55 x 107 m2. Scaling exponents, β, for smaller floe areas range from -0.8 to -0.9 with an average of -0.85 for floe areas ranging from 3 x 106 to 1.55 x 106 m2. The inflection point between the two scaling regimes ranges from 62 x 106 to 151 x 106 m2 and generally moves from larger to smaller floe areas through the summer season. We propose that the two power scaling regimes and the inflection between them are defined during the initial breakup of sea ice solely by the process of fracturing. The distributions of floe size regimes retain their scaling exponents as the floe pack evolves from larger to smaller floe areas from the initial breakup through the summer season, due to scale-independent processes including grinding, crushing, fracture, and melting. The scaling exponents for floe area distribution are in the same range as those reported in previous studies of Antarctic and Arctic floes. A probabilistic model of fragmentation is presented that generates a single power scaling distribution of fragment size.