CS331 Coursework (MATLAB or Python Programming)
Submission Deadline: 31st March 2023 at noon, via Tabula
Submission Format:
o If you use MALTAB, submit 3 scripts (E1.m, E2.m, E3.m) and 3 files
(E1.mlapp, E2.mlapp, E3.mlapp) for app designer.
o If you use Python, submit 3 files (E1.ipynb, E2.ipynb, E3.ipynb) and
make sure they run successfully on Google Colab)
1. Train feedforward neural networks to emulate five 2-input logic gates (AND, OR, NAND, XOR,
IMPLY), respectively. Build a graphical user interface, which a) allows users to select any of these
logic gates from a list box, and b) display the corresponding network diagram (including weights
and biases based on your training) and the truth table of each selected logic gate.
(NB: The network diagram of each logic gate should be drawn using MATLAB, with the edge
weights and biases being generated dynamically from your training results. You are not allowed
to use third-party software, e.g., MS Paintbrush, to draw these diagrams in a static manner.)
(33 marks)
2. Train a neural network to recognise a 7x5 matrix display of digital letters (A-Z, a-z) and numerals
(0-9), as shown in Figure 1. Build a graphical user interface (GUI), which allows users to enter a
string (e.g., “WELCOME”) for testing.
Specifically, you are required to
a) create a textbox to enter a testing string (e.g. “WELCOME”);
b) create a button with caption “Show Original Matrix Display”. When the button is clicked,
visualise the original 7x5 matrix display for each character of your typed string;
c) create a button with caption “Show Noisy Matrix Display”. When the button is clicked,
add 10% noisy pixels at random to each original 7x5 matrix display obtained in step (b)
first and then visualise each perturbed 7x5 matrix display after noise is added.
d) create a button with caption “Predict Noisy Matrix Display”. When the button is clicked,
train your network and label the predicted character for each noisy 7x5 matrix display
obtained in step (c).
(NB: This LED display consists of a 7x5 matrix of lights such that, by turning on or off selected
lights, the required character is displayed. This is a generalisation of the 7-segment LED display.)
(33 marks)
Figure 1 A 7x5 matrix display of digital letters
(A-Z, a-z) and numerals (0-9)
Figure 2 A 7x5 matrix display
7 rows
5 columns
3. Download DBLP dataset (DBLP.mat) from the module page. The DBLP is a co-authorship graph,
consisting of 3823 authors, and their collaborative information. Each node in DBLP denotes an
author. There is an undirected edge between two authors if they co-authored at least one paper.
Build a graphical user interface, which a) shows the names of top-20 most “important” authors
from DBLP in a list box, using PageRank measure; and b) allows users to randomly select any of
them as a query, q, for retrieving the top-10 most “similar” co-authors, x, with respect to query q,
using SimRank score s(x, q), and displaying the top-10 “similar” co-authors, x, in another list box.
(NB: The structure of DBLP.mat is shown in Figure 3, where “A” is a sparse binary adjacency
matrix. A(i,j)=1 if nodes (authors) i and j have co-authored a paper. “authors” is an
“author name” dictionary, where i-th element denotes the author name of node i in the DBLP).
Figure 3 Structure of DBLP.mat
(33 marks)