3789 research outputs found
Sort by
New paradigms of distributed AI for improving 5G-based network systems performance
Get PDFWith the advent of 5G technology, there is an increasing need for efficient and effective
machine learning techniques to support a wide range of applications, from smart cities to
autonomous vehicles. The research question is whether distributed machine learning can
provide a solution to the challenges of large-scale data processing, resource allocation, and
privacy concerns in 5G networks. The thesis examines two main approaches to distributed
machine learning: split learning and federated learning. Split learning enables the separation of model training and data storage between multiple devices, while federated learning
allows for the training of a global model using decentralized data sources. The thesis investigates the performance of these approaches in terms of accuracy, communication overhead,
and privacy preservation. The findings suggest that distributed machine learning can provide a viable solution to the challenges of 5G networks, with split learning and federated
learning techniques showing promising results for spectral efficiency, resource allocation, and
privacy preservation. The thesis concludes with a discussion of future research directions
and potential applications of distributed machine learning in 5G networks.
In this thesis, we investigate four case studies of both 5G network systems and LTE
and Wifi (legacy parts). In chapter3, we implement an asynchronous federated learning
model to predict the RSSI in robot localization indoor and outdoor environments. The
proposed framework provides a good performance in terms of convergence, accuracy, and
overhead reduction. In chapter4, we transfer the deployment of the asynchronous federated
learning framework from the Wifi use case to a part of 5G networks (Network slicing),
where we use the framework to predict the slice type for rapid and automated intelligent
resource allocation. [...
Analysis and simulation of K and Cl in electrostatic precipitator ash and black liquor of an operating kraft recovery boiler
Get PDFNon-process elements such as potassium (K) and chlorine (Cl) are commonly found in the
recovery cycle of kraft mills and are capable of forming highly soluble compounds. These
elements, in high concentrations, can lead to the corrosion of process equipment or to the plugging
of recovery boilers. The behaviour of these elements throughout the liquor cycle is often a direct
result of variables related to the operation of the process. An unusual cyclical trend was observed
in the Cl content of the electrostatic precipitator (ESP) ash produced in the recovery boiler of a
Canadian kraft mill. Analysis of the ash and liquor samples during a 390-day sampling period
found that the Cl enrichment factor (EFCl) varied significantly between 1.2 to 3.9, whereas the EFK
stayed fairly constant at ~ 1.4. [...
Development of a consistent cubic equation of state for the calculation of the phase behaviour and thermodynamic properties of pure components
Get PDFThis study analyzes two cubic equations of state (CEoS), the Redlich-Kwong (RK) and PengRobinson (PR) CEoS, to calculate the properties of 151 pure components. The components were
grouped by their intermolecular interactions (polar, non-polar, hydrogen bonding) and their
saturation pressure (P
sat), enthalpy of vaporization (∆Hv), and saturated liquid heat capacity (Cp
sat)
were calculated using a unique combination of the PR and RK CEoS paired with either the Twu or
Soave α-functions. Outliers were eliminated using a quantile regression algorithm; additionally, a
set of constraints proposed by Le Guennec et al.(Le Guennec et al. 2016) was applied to ensure
that thermodynamic consistency was enforced during optimization. The four models tested were
named PR-Twu, PR-Soave, RK-Twu, and RK-Soave, based on the equation of state and alpha
function used. The PR-Twu and RK-Twu models predict P
sat
, ∆Hv, and Cp
sat more accurately than
the PR-Soave and RK-Soave models. [...
Synthesized emulsifiers from kraft lignin and the role of mixing in oil-in-water emulsions
Get PDFOil-in-water emulsions are widely used in various industrial applications such as food,
pharmaceuticals, and cosmetics. An oil-in-water emulsion is a colloidal system in which oil
droplets are dispersed in water with the help of emulsifiers. The emulsifiers can stabilize the oil
droplets in the water phase, preventing them from coalescing and being separated from the water,
resulting in a stable mixture that can be used in various industrial processes. Lignin
macromolecules, derived from renewable biomass resources, have gained extensive interest during
the past decade as a sustainable substitute for oil-based synthetic materials. One of the goals of
this dissertation was to formulate and develop a lignin-based emulsifier that is renewable,
biodegradable, and non-toxic from the molecular level to the macroscopic level. To obtain the
desired physicochemical properties, chemical modifications were conducted. We utilized solventfree reactions to synthesize sulfo-alkylated lignin-based emulsifiers in a facile green process. Oilin-water emulsions were formulated in the presence of emulsifiers, and a vertical scan analyzer
was used to monitor their stability to show their potential for stabilizing emulsions. [...
A multi-substrate strontium isotope baseline for the Promontory Caves, Utah: implications for studies of ancient bison migration
Get PDFThe purpose of this thesis is to provide a framework for evaluating bison mobility in the
eastern Great Basin during the thirteenth-century using strontium (Sr) isotope analysis. The
Promontory Caves, Utah (42BO1 and 42BO2) were occupied for a relatively short period
(A.D.1250-1290) but have a rich record of incredibly well-preserved organic remains including a
high abundance of bison remains indicating that bison were a key prey species. Previous research
indicates a decline in the local bison population which may have triggered a push for ancient
people to navigate the landscape to shift their home (or seasonally-used territory). One possible
site that the Promontory people visited is West Fork Rock Creek (WFRC) (10-Oa-275), Idaho.
There is evidence that WFRC was visited by Promontory people as they were hunting bison in
the late thirteenth-century. [...
Cannabis use among women experiencing menstrual cycle distress
Get PDFWomen experiencing menstrual cycle distress might use cannabis for self-medicative purposes.
The purpose of this study was to determine whether cannabis use frequency and mode of intake
were associated with the type of symptoms experienced (i.e., affective, physiological),
throughout the menstrual cycle (i.e., menstrual phase premenstrual phase, and during the
remainder of the menstrual cycle). Of secondary interest, we explored women’s awareness of
cannabinoid content. We additionally explored cannabis use and menstrual cycle distress
symptoms across the phases of the menstrual cycle. Participants (N = 147) were recruited
through a course credit system, and community and online advertisements. Participants selfreported on their most recent menstrual cycle and cannabis use using a web-based survey.
Participants were categorized as current cannabis users (n = 82) versus non-users (n = 65).
Results indicated no significant evidence to suggest associations between cannabis use and
menstrual cycle distress symptoms. Cannabis users who were aware of the cannabinoid content
of their cannabis products used cannabis more days than those unaware of the cannabinoid
content. Cannabis users reported experiencing more pain during the menstrual and premenstrual
phases, compared to non-users. This difference was not seen for the remainder of the cycle
phase. Findings from this study expand on previous research regarding pain severity across the
phases of the menstrual cycle by specifically comparing cannabis users and non-users
Full-duplex MU-MIMO systems under the effects of non-ideal transceivers: performance analysis and power allocation optimization
Get PDFModern Technologies, particularly connectivity, increasingly support many facets of everyday life. The next generation of wireless communication systems aims to provide new
advanced services and support new demands. These services are required to serve a massive number of devices and achieve higher spectral and energy efficiency, ultra-low latency,
and reliable communication. The research community around the globe is still working on
finding novel technologies to meet these requirements. Full duplex (FD) communications
have been recognized as one of the promising wireless transmission candidates and gamechangers for the future of wireless communication and networking technologies, thanks to
their ability to greatly improve spectral efficiency (SE) and dramatically enhance energy
efficiency (EE). In this thesis, first, the influence of hardware impairment (HWI) on singleinput single-output (SISO) FD access point (AP) is studied. More precisely, the SE and
EE when the system’s terminals have impaired transceivers are analyzed. Optimization
problem for EE maximization is formulated to fulfill quality of service (QoS) and power
budget constraints. An algorithm to solve the optimization problem by using the fractional
programming theory and Karush–Kuhn–Tucker (KKT) conditions technique is proposed. [...
Multi-objective resource optimization in space-aerial-ground-sea integrated networks
Get PDFSpace-air-ground-sea integrated (SAGSI) networks are envisioned to connect satellite, aerial, ground,
and sea networks to provide connectivity everywhere and all the time in sixth-generation (6G) networks. However, the success of SAGSI networks is constrained by several challenges including
resource optimization when the users have diverse requirements and applications. We present a
comprehensive review of SAGSI networks from a resource optimization perspective. We discuss
use case scenarios and possible applications of SAGSI networks. The resource optimization discussion considers the challenges associated with SAGSI networks. In our review, we categorized
resource optimization techniques based on throughput and capacity maximization, delay minimization, energy consumption, task offloading, task scheduling, resource allocation or utilization,
network operation cost, outage probability, and the average age of information, joint optimization (data rate difference, storage or caching, CPU cycle frequency), the overall performance of
network and performance degradation, software-defined networking, and intelligent surveillance
and relay communication. We then formulate a mathematical framework for maximizing energy
efficiency, resource utilization, and user association. We optimize user association while satisfying
the constraints of transmit power, data rate, and user association with priority. The binary decision
variable is used to associate users with system resources. Since the decision variable is binary and
constraints are linear, the formulated problem is a binary linear programming problem. Based on
our formulated framework, we simulate and analyze the performance of three different algorithms
(branch and bound algorithm, interior point method, and barrier simplex algorithm) and compare
the results. Simulation results show that the branch and bound algorithm shows the best results,
so this is our benchmark algorithm. The complexity of branch and bound increases exponentially
as the number of users and stations increases in the SAGSI network. We got comparable results
for the interior point method and barrier simplex algorithm to the benchmark algorithm with low
complexity. Finally, we discuss future research directions and challenges of resource optimization
in SAGSI networks
Do the stigma of mental illness and the stigma of non-suicidal self-injury intersect?
Get PDFBackground: It is well-known that individuals with a mental illness (MI) are highly stigmatized.
Oftentimes, the public views these individuals as blameworthy, and this often leads to
discrimination, segregation, and avoidance of those with a MI. Due to high rates of
stigmatization, individuals with MI often do not seek help for their issues. Stigmatization can
also occur within the scope of non-suicidal self-injury (NSSI), which refers to the deliberate
damage of one’s tissue, without suicidal intent (e.g., cutting one’s skin). Often, MI and NSSI cooccur, however currently no research exists as to how the stigma of these two entities intersects.
Purpose: To investigate whether the stigmatization of MI and NSSI intersect. In other words, is
a person with a MI who engages in NSSI more stigmatized than one who does not self-harm? It
was hypothesized that a person who has a MI and self-harms will be more stigmatized than an
individual who has a MI, but does not engage in NSSI. Secondarily, is also hypothesized that
stigma will manifest in different ways, depending on the disorder described. Based on the
previous literature, it is likely that borderline personality disorder (BPD) will be more
stigmatized than both post-traumatic stress disorder (PTSD) and depression (DEP), and that
PTSD will be more stigmatized than DEP. [...
Efficient path planning and battery management for electric vehicles
Get PDFThe rapid advancement in battery technology has brought electric vehicles (EVs) into
reality, and the increasing adoption of autonomous electric vehicles (AEVs) has presented
significant challenges. Existing research in the realm of IoT has extensively explored EV
transportation systems, focusing on aspects like routing, energy management, and grid
system equilibrium. In this context, this thesis readdresses the challenge of determining the
fastest route for AEVs considering the battery charging time.
Diverging from the current state-of-the-art, our work delves into the prospect of not
only minimizing travel time but also maximizing battery life for the optimal utilization of
electric vehicles. We commence by formalizing the problem of ”Efficient Path Planning
and Battery Management for Electric Vehicles” as a mixed integer linear programming
(MILP) model, thereby deriving its optimal solutions mathematically. Given the inherent
complexity of the optimization model, we introduce a range of heuristic algorithms designed
to address the problem at scale. Furthermore, this problem is similar to the traveling
salesman problem(TSL), which means it has an NP-hard nature. [...