Module Database Search
MODULE DESCRIPTOR | |||
---|---|---|---|
Module Title | |||
Deep Learning | |||
Reference | CM3604 | Version | 3 |
Created | February 2024 | SCQF Level | SCQF 9 |
Approved | July 2020 | SCQF Points | 15 |
Amended | April 2024 | ECTS Points | 7.5 |
Aims of Module | |||
---|---|---|---|
To provide theoretical and practical experience to train, test and deploy deep learning models using production ready state-of-the-art deep learning frameworks. |
Learning Outcomes for Module | |
---|---|
On completion of this module, students are expected to be able to: | |
1 | Review a range of deep learning architectures for a given learning problem. |
2 | Experiment with a deep neural net architecture for a range of problems (both structured and unstructured). |
3 | Formulate a neural model by selecting relevant functions and parameters using a state-of-the-art python framework. |
4 | Experiment with transfer learning to enhance deep learning models for cross-domain applications. |
Indicative Module Content |
---|
Deep learning framework: Any suitable framework such as Tensorflow, Keras, PyTorch etc. Supervised architectures: Feedforward nets, Recurrent nets, Convolutional neural nets for images and LSTMs for sequence data (textual content). Unsupervised architectures: autoencoders , variational autoencoders, generative adversarial networks; Model weighting mechanisms such as hard and soft attention as well as self attention. Meta learners: deep metric learners like siamese networks, matching networks, model agnostic learners. Zero-shot and few-shot learning paradigms and importance of transfer learning for model adaptation. Role of ethics and transparency in deep learning. |
Module Delivery |
---|
The module delivery will consist of a series of lectures and tutorials; both online and on-campus blended delivery of the module. The theoretical concepts and main principles, algorithms will be introduced during the lecture and the students will be provided with exercises during the lecture to apply and test their theoretical knowledge in-class. The tutorial sessions will consist of practical exercises to apply the theoretical principles to real world problems. |
Indicative Student Workload | Full Time | Part Time |
---|---|---|
Contact Hours | 48 | N/A |
Non-Contact Hours | 102 | N/A |
Placement/Work-Based Learning Experience [Notional] Hours | N/A | N/A |
TOTAL | 150 | N/A |
Actual Placement hours for professional, statutory or regulatory body |   |   |
ASSESSMENT PLAN | |||||
---|---|---|---|---|---|
If a major/minor model is used and box is ticked, % weightings below are indicative only. | |||||
Component 1 | |||||
Type: | Coursework | Weighting: | 100% | Outcomes Assessed: | 1, 2, 3, 4 |
Description: | Individual coursework covering all learning outcomes. |
MODULE PERFORMANCE DESCRIPTOR | |
---|---|
Explanatory Text | |
The calculation of the overall grade for this module is based on 100% weighting of C1. An overall minimum grade of D is required to pass the module. | |
Module Grade | Minimum Requirements to achieve Module Grade: |
A | The student needs to achieve an A in C1. |
B | The student needs to achieve a B in C1. |
C | The student needs to achieve a C in C1. |
D | The student needs to achieve a D in C1. |
E | The student needs to achieve an E in C1. |
F | The student needs to achieve an F in C1. |
NS | Non-submission of work by published deadline or non-attendance for examination |
Module Requirements | |
---|---|
Prerequisites for Module | CM1606, CM2602, CM2604 and CM2607 or equivalents. |
Corequisites for module | None. |
Precluded Modules | None. |
INDICATIVE BIBLIOGRAPHY | |
---|---|
1 | Geron, A. 2019. Hands-on Machine Learning with Scikit-Learn and TensorFlow. 2nd ed. O’Reilly. |
2 | Ameisen, E. 2020. Building Machine Learning Powered Applications: Going from Idea to Product. O’Reilly. |
3 | Goodfellow, I. , Bengio, Y. and Courville, A. 2016. Deep Learning. MIT Press |
4 | Bishop, C. 2007. Pattern Recognition and Machine Learning. Springer-Verlag. |