CSC8637 : Deep Learning
- Offered for Year: 2024/25
- Module Leader(s): Dr Stephen McGough
- Lecturer: Dr Huizhi Liang
- Owning School: Computing
- Teaching Location: Newcastle City Campus
Semesters
Your programme is made up of credits, the total differs on programme to programme.
| Semester 2 Credit Value: | 10 |
| ECTS Credits: | 5.0 |
| European Credit Transfer System | |
Aims
Deep Learning is a sub-field within the area of Machine Learning which has attracted significant interest during recent years most notably because of its ability to outperform humans on many tasks which previously were considered to be too hard to perform on a computer. Deep Learning is all around us and used on a daily basis. If you talk to your phone or a home assistant, it’s Deep Learning which is converting your speech into text. If you search for friends on social media by giving a picture of them, then this is using Deep Learning. Even the adverts you see when surfing the web are likely to have been chosen for you by Deep Learning. Deep Learning is also going to be one of the key technologies for future developments such as self-driving cars and autonomous robots.
This module will introduce students to the area of Deep Learning, the different technologies available along with the myriad of application areas where it can be applied. The aim is to make students fluent in the approach of Deep Learning which will allow them to build up skills that will enable them to apply these skills for applications in companies, academia or the third sector.
Deep Learning is a powerful tool which, like all powerful tools, be used for good or bad. Hence a running theme within this module will be the ethical use of Deep Learning along with ensuring that students are aware of the biases which can be present and approaches to minimising such bias. Students will be made aware of good practices for Deep Learning and how to report their results in a fair and honest manner.
Outline Of Syllabus
Material will cover such topics as:
• Multi-Level Perceptrons (MLP)
• Backpropagation, gradient decent and optimisation
• Convolutional Neural Networks (CNN)
• Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU)
• Natural Language Processing (NLP)
• Generative Models, e.g. Generative Adversarial Networks (GAN), and Variational Autoencoders (VAE)
• Better training of neural networks, e.g. Preventing Overfitting, Dropout, Transfer Learning, Data
Augmentation
• Ethics and Challenges of Deep Learning, Adversarial Examples, and good practices in Deep Learning
• Interoperability and Explainability of Deep Learning
• Embedding
• Deep Learning on Graphs, GraphCNN
Teaching Methods
Teaching Activities
| Category | Activity | Number | Length | Student Hours | Comment |
|---|---|---|---|---|---|
| Guided Independent Study | Assessment preparation and completion | 1 | 1:00 | 1:00 | Oral exam |
| Guided Independent Study | Assessment preparation and completion | 2 | 1:00 | 2:00 | Preparation for oral exam |
| Guided Independent Study | Assessment preparation and completion | 30 | 1:00 | 30:00 | Coursework project |
| Scheduled Learning And Teaching Activities | Lecture | 8 | 1:00 | 8:00 | Lectures (online, non-synchronous) |
| Scheduled Learning And Teaching Activities | Practical | 1 | 3:00 | 3:00 | Practicals |
| Guided Independent Study | Independent study | 12 | 1:00 | 12:00 | Background reading |
| Guided Independent Study | Independent study | 5 | 1:00 | 5:00 | Seminar preparation |
| Guided Independent Study | Independent study | 18 | 1:00 | 18:00 | Lecture follow-up |
| Scheduled Learning And Teaching Activities | Scheduled on-line contact time | 7 | 3:00 | 21:00 | Seminars of current research (synchronous, PiP) |
| Total | 100:00 |
Teaching Rationale And Relationship
Lectures explain the underpinning principles for the module and technologies that support deep learning and exploratory data analysis. Lectures are complemented by supervised practical sessions to guide the application of these principles using suitable computational tools. The practical work builds up experience working with a computational toolset that is used to complete a substantive project working with data from a real-world context.
Assessment Methods
The format of resits will be determined by the Board of Examiners
Other Assessment
| Description | Semester | When Set | Percentage | Comment |
|---|---|---|---|---|
| Report | 2 | M | 100 | Extended technical project Word count: Up to 2,000 words |
Zero Weighted Pass/Fail Assessments
| Description | When Set | Comment |
|---|---|---|
| Oral Examination | M | Presentation and demonstration of the methods and results from the coursework project. Presentation length: 15 mins |
Formative Assessments
Formative Assessment is an assessment which develops your skills in being assessed, allows for you to receive feedback, and prepares you for being assessed. However, it does not count to your final mark.
| Description | Semester | When Set | Comment |
|---|---|---|---|
| Prof skill assessmnt | 2 | M | Having read the papers, reflect on their understanding of the papers and how they can deepen their understanding from here. |
Assessment Rationale And Relationship
The formative assessment are quizzes done in class to aid learning. The report tests the students’ ability to apply deep learning techniques in a reproducible manner, using effective tools and methods to solve a real-world challenge. The presentation assesses the students’ ability to communicate their findings and approach. The formative assessment facilitates a reflective learning.
Reading Lists
Timetable
- Timetable Website: www.ncl.ac.uk/timetable/
- CSC8637's Timetable