MS4H20 - Advanced Tools and Techniques for Data Science 27 Jul 2022 - 31 Aug 2026 | Version 1

Associated Module Information

Module Code: MS4H20
Module Title: Advanced Tools and Techniques for Data Science
Faculty: Faculty of Computing, Engineering and Science
Faculty Group: Computing and Mathematics
Faculty Sub Group: Maths
Module Leader: Ieuan Griffiths
Module Team: Rebecca Peters, Samuel Jobbins, Angelica Pachon, Stephanie Perkins
First Intended Intake: Final Year of Intake:
Date Closed:
Credit Value: 10 Credit Level: 7
Language: English
Percentage of Module Taught in Welsh: 0
Equivalent Module:
HECOS codes: 100359 - artificial intelligence 100992 - machine learning
HECOS Code Weighting: 50 50

Document Version Information

Version 1
Valid From 27 Jul 2022
Valid To 31 Aug 2026

Module Aims

To provide an understanding of Deep Neural Networks, the underlying mathematics behind them, and the techniques for using them as a powerful machine learning tool.

To select appropriate architectures of Artificial Neural Networks, how optimisation algorithms work, and how to identify hyperparameters in order toto improve Networks performance.?Furthermore, the module looks to provide a foundational understanding of model deployment within a cloud computing environment.

Content Summary

This module focuses on the understanding and implementation of Artificial Neural Networks (ANN). It will start by reviewing classification algorithms studied during the Machine Learning module, as 'black box' models. It will then discuss limits of linear classifiers and how to tackle problems using multilayer feedforward networks. A thorough presentation of stochastic gradient descent and backpropagation algorithms will guide the students in the implementation of a simple neural network from scratch. The use of cutting-edge libraries to implement bigger architectures will lead to the creation of an image classifier. The module will conclude with the study and implementation of Convolutional Neural Networks (CNN).?Finally, students will explore cutting edge Big Data technologies and Cloud Computing Solutions, such as Microsoft Azure’s ML Platform/ Amazon Web Services. In doing so, students will understand the limitations of executing complex algorithms on local system hardware, and develop an understanding of model deployment in a cloud computing environment.

Learning and Teaching Methods

Activity Type Hours
Lecture 8
Seminar 8
Tutorial 8
Independent Study 40
Directed Study 36
Total Hours Selected 100

Learning Outcomes

# Learning Outcome
LO1 To understand the concepts of Artificial Neural Network, loss function, optimisation algorithm, and how the backpropagation algorithm works.
LO2 To implement from scratch simple Artificial Neural Networks and more sophisticated architectures using cutting-edge machine learning software libraries.
LO3 To obtain hands on experience with model deployment in a cloud computing environment, utilising industry leading solution.

Module Requisites

N/A

Assessment Criteria

Assessment Category Assessment Type Description Duration Word Count Weight (%) Best of? Pass Mark
Asynchronous Assessment Practical Coursework 1 (Asynch) Analyse a data source using suitable deep learning techniques and report appropriate conclusions from the results of the analysis and deployment. 0 2000 100 No 40

Assessment Matrix

Assessment Type Learning Outcomes
LO1 LO2 LO3
Practical Coursework 1 (Asynch)

Reading List

Geron, A. (2019) Hands-on Machine Learning with Scikt-Learn, Keras & TensorFlow. 2nd edn. O’Reilly.

Goodfellow, I., Bengio, Y., Courville, A. (2017) Deep Learning (Adaptive Computation and Machine Learning Series) MIT Press.

Raschka, S., Mirjalili, V. (2019) Python Machine Learning. 3rd edn. Packt Publishing.