Add further Overleaf changes

project-report/src/latex/analysis.tex

@@ -115,12 +115,20 @@ \subsubsection{How Artificial Neural Networks learn}
     \item Having a 'Vanishing Gradient', where the gradient value decreases to a very small value or zero, resulting in a lack of updating values during training.
 \end{itemize}
 
-\vspace{5mm}
+\pagebreak
+
 \subsection{Theory Behind Deep Artificial Neural Networks}
 \vspace{5mm}
 
 \subsubsection{Network Architecture and Training}
 
+\begin{figure}[h!]
+\centering
+\includegraphics[width=1\textwidth]{./project-report/src/images/deep-ann-diagram-2.png}
+\caption{This shows an Artificial Neural Network with multiple hidden layers and is known as a Deep Neural Network.}
+\label{fig:abstract-network}
+\end{figure}
+
 Figure \ref{fig:abstract-network} below shows a simplified view of an Artificial Neural Network with multiple hidden layers, known as a Deep Neural Network, where:
 
 \pagebreak

project-report/src/latex/evaluation.tex

@@ -7,6 +7,29 @@
 \clearpage
 \section{Evaluation}
 
+\subsection{Third Party Feedback}
+
+I demonstrated the final version of my program to the same third party that I interviewed in the analysis, and their response is shown below:
+
+\vspace{5mm}
+
+"In my opinion, Max has definitely met the primary and secondary goals of this project. Firstly, and most importantly, he has researched and implemented, from first 
+principles, an Artificial Neural Network that is flexible and abstracted to the point that it can tackle a range of problems. Max started the analysis for this 
+project from a very theoretical and mathematical point of view before implementing code which has allowed him to extend its implementation to a range of datasets 
+from the XOR problem to image analysis.
+
+I was particularly impressed at the level of analysis he undertook into how Artificial Neural Networks work and the impact that different kinds of design decisions can have on 
+implementation. He took on board suggestions to explore different types of transfer functions such as the ReLu function to increase the speed of training and it was nice 
+to see comparative studies of this and other techniques. It was also great to see the ability to save and load trained models which has allowed him to train models 
+on a desktop PC equipped with a graphics card to be utilized on a lower power laptop.
+
+The analysis section exploring the impact on both learning rates and epoch count was very nice to see, as well as the identification of an optimal learning rate 
+suitable for the image dataset he was working with.
+
+In summary, it was fantastic to see a true maths-to-code example of implementing Artificial Neural Networks that didn't rely on the use of external AI libraries. I 
+am certain he has learned a great deal regarding the fundamental properties and limitations of Artificial Neural Networks. I was also impressed by the usage of 
+software engineering tools such as GitHub and Jupyter Notebook throughout the project."
+
 \subsection{Project Objectives Evaluation}
 
 \subsubsection{Project Objectives}
@@ -59,28 +82,43 @@ \subsubsection{Project Objective Evaluations}
     \hline
 \end{tabular}
 
-\subsection{Third Party Feedback}
-
-I demonstrated the final version of my program to the same third party that I interviewed in the analysis, and their response is shown below:
-
-\vspace{5mm}
-
-"In my opinion, Max has definitely met the primary and secondary goals of this project. Firstly, and most importantly, he has researched and implemented, from first 
-principles, an Artificial Neural Network that is flexible and abstracted to the point that it can tackle a range of problems. Max started the analysis for this 
-project from a very theoretical and mathematical point of view before implementing code which has allowed him to extend its implementation to a range of datasets 
-from the XOR problem to image analysis.
-
-I was particularly impressed at the level of analysis he undertook into how Artificial Neural Networks work and the impact that different kinds of design decisions can have on 
-implementation. He took on board suggestions to explore different types of transfer functions such as the ReLu function to increase the speed of training and it was nice 
-to see comparative studies of this and other techniques. It was also great to see the ability to save and load trained models which has allowed him to train models 
-on a desktop PC equipped with a graphics card to be utilized on a lower power laptop.
-
-The analysis section exploring the impact on both learning rates and epoch count was very nice to see, as well as the identification of an optimal learning rate 
-suitable for the image dataset he was working with.
-
-In summary, it was fantastic to see a true maths-to-code example of implementing Artificial Neural Networks that didn't rely on the use of external AI libraries. I 
-am certain he has learned a great deal regarding the fundamental properties and limitations of Artificial Neural Networks. I was also impressed by the usage of 
-software engineering tools such as GitHub and Jupyter Notebook throughout the project."
+\subsection{Requirements Evaluation}
+
+\begin{tabular}{|p{0.03\linewidth}|p{0.69\linewidth}|p{0.12\linewidth}|p{0.16\linewidth}|}
+	  \hline
+      \textbf{ID} & \textbf{Description} & \textbf{Status} & \textbf{3rd Party Evaluation} \\
+      \hline
+	  1 & Learn how Artificial Neural Networks work & Fully met & Fully met \\
+      \hline
+	  2 & Develop Artificial Neural Networks from first principles & & \\
+      \hline
+	  2.1 & Provide utilities for creating Artificial Neural Networks & Fully met & Fully met \\
+      \hline
+	  2.2 & Allow for the saving and loading of trained models' weights and biases & Fully met & Fully met \\
+      \hline
+	  2.3 & Allow use of Graphics Cards for faster training & Fully met & Fully met \\
+      \hline
+      3 & Implement the Artificial Neural Networks on image datasets & & \\
+      \hline
+	  3.1 & Allow input of unique hyper-parameters & Fully met & Fully met \\
+      \hline
+	  3.2 & Allow unique datasets and train dataset size to be loaded & Fully met & Fully met \\
+      \hline
+	  4 & Use a database to store a model's features and the location of its weights and biases & Fully met & Fully met \\
+      \hline
+      5 & Develop a Graphical User Interface & & \\
+      \hline
+	  5.1 & Provide controls for hyper-parameters of models & Fully met & Fully met \\
+      \hline
+	  5.2 & Display details of models' training & Fully met & Fully met \\
+      \hline
+	  5.3 & Display the results of each model's predictions & Fully met & Fully met \\
+      \hline
+	  5.4 & Allow for the saving of trained models & Fully met & Fully met \\
+      \hline
+	  5.5 & Allow for the loading of saved trained models & Fully met & Fully met \\
+      \hline
+\end{tabular}
 
 \subsection{Future Improvements}
 
@@ -96,4 +134,4 @@ \subsection{Future Improvements}
     \item Utilising a standardized file format for storing trained Artificial Neural Networks, so that they can be integrated with other machine learning libraries.
 \end{itemize}
 
-\end{document}
+\end{document}