GRA 4162 Deep Learning and Explainable AI
GRA 4162 Deep Learning and Explainable AI
Deep learning models have achieved state-of-the-art performance in tasks such as image classification, representation learning, and data generation. This course explores various deep learning architectures, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), transformers, and autoencoders; as well as probabilistic graphical models and variational autoencoders (VAEs). Students will gain hands-on experience in implementing and training deep learning models using TensorFlow.
Additionally, as interpretability of the results is crucial in many real-world applications, the course introduces principles and techniques of Explainable Artificial Intelligence (XAI), equipping students with the skills to make deep learning models more transparent and interpretable.
By the end of the course, the students will be able to
- Describe the fundamental concepts of deep learning and neural networks.
- Explain the differences between the various deep learning architectures and their learning algorithms.
- Apply key principles and techniques of Explainable AI for model interpretability.
- Use generative models to deal with limited amount of data.
By the end of the course, the students will be able to
- Implement, train, and evaluate deep learning models using TensorFlow.
- Choose appropriate methodology for a given learning problem and given data set.
- Use XAI methods to interpret and explain deep learning models and their predictions.
By the end of the course, the students will be able to:
- Critically assess the strengths and limitations of deep learning models.
- Work effectively in teams to develop and analyze deep learning models.
- Communicate technical findings and insights clearly to both technical and non-technical audiences.
- Apply ethical considerations in AI, particularly regarding model transparency and bias.
- Stay updated with the advancements in deep learning and XAI through academic literature and industry trends.
To address the learning outcomes listed above, the course content covers the following topics:
· Introduction to deep learning - history, positioning in AI, applications, and recent developments and challenges
· Feedforward neural networks – backpropagation and gradient descent
· Image processing with CNNs
· Sequential data modelling with RNNs and Transformers
· Representation learning with Autoencoders
· Generative modelling with VAEs and probabilistic graphical models
· General principles and a choice of techniques of Explainable AI
Learning activities will combine lectures and laboratory sessions. Students are expected to prepare for lectures by reading assigned material and actively participate in discussion of the lecture topics. In addition, students will work on a practical assigment, which will be answered in groups of 1-3 students. The software tool is Python, with emphasis on TensorFlow.
Please note that while attendance is not compulsory in all courses, it is the student’s own responsibility to obtain any information provided in class.
All parts of the assessment must be passed in order to get a grade in the course.
All courses in the Masters programme will assume that students have fulfilled the admission requirements for the programme. In addition, courses in second, third and/or fourth semester can have spesific prerequisites and will assume that students have followed normal study progression. For double degree and exchange students, please note that equivalent courses are accepted.
GRA4153, GRA4160, GRA4152, and 15 ECTS undergraduate credits in mathematics or statistics/data science, and programming. The graduate courses can be compensated with 18 ECTS in mathematics or statistics/data science, and programming.
| Assessments |
|---|
Exam category: School Exam Form of assessment: Written School Exam - pen and paper Exam/hand-in semester: First Semester Weight: 40 Grouping: Individual Support materials:
Duration: 3 Hour(s) Comment: . Exam code: GRA 41621 Grading scale: ECTS Resit: Examination when next scheduled course |
Exam category: Submission Form of assessment: Submission other than PDF Exam/hand-in semester: First Semester Weight: 60 Grouping: Group (1 - 3) Duration: 3 Week(s) Comment: Written assignment. Exam code: GRA 41622 Grading scale: ECTS Resit: Examination when next scheduled course |
All exams must be passed to get a grade in this course.
| Activity | Duration | Comment |
|---|---|---|
Teaching | 36 Hour(s) | 3 hours per week (2 synchronous + 1 asynchronous) for 12 weeks |
Seminar groups | 14 Hour(s) | 2 hours per week for 7 weeks |
Student's own work with learning resources | 110 Hour(s) |
A course of 1 ECTS credit corresponds to a workload of 26-30 hours. Therefore a course of 6 ECTS credits corresponds to a workload of at least 160 hours.
