FINAL YEAR THESIS PRESENTATION SLIDES 2024.ppt

The Thesis presentation slides for Masters in Information Technology

1. BY
NAME: MUMUNI RASHEED OLUBUNMI
MATRIC NO: NOU222051625
COURSE CODE: CIT 899
COURSE TITLE: RESEARCH PROJECT
UNDER THE SUPERVISION OF DR. ADEGOKE OJENIYI
NATIONAL OPEN UNIVERSITY OF NIGERIA
ABUJA, NIGERIA.
THESIS PRESENTATION ON

2. Introduction to the Study
Problem Statement and Objectives
Review of Related Studies
Methodology
Results and Findings
Conclusion
Summary

3. The wireless and smart technologies like 4G, 5G, IoT, IoE,
and the upcoming 6G networks has greatly enhanced
connectivity for the vast number of smart or intelligent
devices and machines.
The task is how these devices and machines intelligently learn
experience, independently enhance network structures and
quickly make decisions with the slightest human intervention.
Executing intelligent networks gives rise to a lot of challenges
based on amount of data generated by the number of smart
devices like collecting, accessing and processing the massive
amount of data.
This thesis focuses on network optimization and resource
allocation based on GNN to solve key challenges in wireless
network with a case study of NOUN, Abeokuta Study Centre.

4. In NOUN, Abeokuta Study Centre, does communication and resource sharing
between devices actually take place? The questions as detailed in this study are
as below:
i.How to examine and study the influence of DL on the behavioral analysis of
devices and their impact on wireless network?
ii.How to study the effect of DL on identification of unjustifiable signal strength
and overcrowding points with latency in wireless network?
iii.How to create positive association between application types availability,
bandwidth requirements and interoperability in wireless network for resource
allocation?
iv.The thesis aims are summarized as follows:
v.To implement a graph based DL structure that embeds the network topologies to
estimate the resource allocation difficulties in wireless network.
vi.To use the graph based DL methodology to solve complex optimization
difficulties.
vii.To implement a graph based DL methodology to solve resource sharing
allocation problems.

5. DL is a subset of ML technique in AI that train computers to process data
in a way that is stimulated by the human brain.
 AI attempts to train
computers to think and learn as humans do and DL technology drives
many AI applications used in everyday products.
DL models are computer files that data scientists have trained to perform
tasks using an algorithm or a predefined set of steps.
Supervised learning is a technique that the training of data is often brand-
named by a data scientist in the planning phase, before being used to
train and test the model.
Unsupervised learning is a technique that does not require labeled data
for training and often used to recognize patterns and developments in raw
datasets.
A Graph is the arrangement of data structure that comprises nodes and
edges. A node can be a person, place, or thing, and the edges express the
relationship between nodes. The edges can be directed and undirected
based on directional needs.
 

6. GNN are exceptional types of neural networks capable of
working with a graph data and they are used in forecasting
nodes, edges, and graph-based tasks.
 
Due to the similarities between wireless network topologies and
graphs, GNN can be used to solve the tricky resource allocation
problems using algorithms. The integration of the graph theory
and the DL methods has emerged as a vital research topic.
GNN was introduced when CNN failed to accomplish optimal
results due to the haphazard size of the graph and complex
structure.
 
In this thesis, the wireless network is modeled as fully connected
graph consisting of nodes and edges. Since the functions of the
communication and interference links work in the opposite
ways, it is rational to treat the communication links and
interference links as the nodes and the edges, respectively.

7. In this study, a graph-based DL represented by GNN is
proposed for making decisions in a series of problems. It is
used to address problems in a supervised manner.
Countless resource distribution problems in wireless network
can be articulated as an optimization problem, which is
typically problematic to find the optimal solutions. A universal
design of this kind of problems can be engraved as follows:
min f(x)
subject to x = x
i
,
∈ℕ
g
n
(x) ≤ 0,
This requires adequate labeled training samples that involves
an iterative procedure and the iteration is divided into two
phases: (i) initially produce random samples according to a
definite probability distribution (ii) reduce this distribution and
a target distribution then update the distribution centered on
the data to produce better samples in the next iteration.

8. The universal constrained resource sharing problem algorithm
adopted are summarized below:
Algorithm for Optimization Problem:
Prepare the Bernoulli probability vector B
(0)
={B
(0)
i
}and iteration
index t = 1.
Produce a random large number of S samples where {x
j
}
j=1
according to the probability B
(t−1)
, where the i-th element of x
j
is
denoted by x
j
i
.
Create expectations of any impracticable samples into
practicable samples.
Assemble {x
j
}
j=1
in an ascending order as {x
σj
}
σj=1
with respect to
measure values calculated by f(x
j
).
Handpicked the best samples from {x
σj
}
σj=1
and update the
probability vector B
t
with B
t
i
= ∑
σj=1
x
i
σj
if B
t
does not converge
to a binary vector then set t = t + 1 and go to step 2 else x = B
t
.
end if

9. The communication graph of NOUN Abeokuta Study Centre network is
presented, which consists of 11 nodes and 14 edges.

11. The interpretation and connectivity exploration provides valuable
comprehensions into how dissimilar applications affect the QoS in the
wireless network which is then used to optimize network configurations
and advance user experience. This involves exploring a wealth of data
encompassing various aspects of wireless network resource allocation,
including:
a.Application Types: This gain insight into how different applications,
from high definition video calls, streaming, online gaming and so on,
demand and receive network resources.
b.Signal Strength: This give an understanding of how signal strength
impacts resource allocation decisions and quality of service.
c.Latency: This is to discover the delicate balance between low-latency
requirements and resource availability.
d.Bandwidth Requirements: This dive into the diverse bandwidth needs
of applications and their influence on allocation percentages.
e.Resource Allocation: This explores the core of dynamic resource
allocation, where percentages reflect the deep learning decisions that
ensure optimal network performance.

12. 
The incorporation of deep learning in this thesis has
brought numerous benefits, including:
a.Improved network performance through intelligent
traffic management and anomaly detection.
b.Enhanced security through behavioral
authentication, real-time threat detection, and
adaptive firewalls.
c.Intelligent network management with automated
resource allocation and network optimization.
d.Embracing these advancements will undoubtedly
lead to enhanced network performance, improved
user experiences, and increased competitive
advantage in the digital era.