GRA 4162 Deep Learning and Explainable AI

GRA 4162 Deep Learning and Explainable AI

Course code: 
GRA 4162
Department: 
Data Science and Analytics
Credits: 
6
Course coordinator: 
Magdalena Ivanovska
Course name in Norwegian: 
Deep Learning and Explainable AI
Product category: 
Master
Portfolio: 
MSc in Data Science for Business
Semester: 
2024 Autumn
Active status: 
Active
Level of study: 
Master
Teaching language: 
English
Course type: 
One semester
Introduction

The performance of deep learning models in tasks such as image classification, representation learning or data generation has reached state-of-the-art results in recent years.

In this course we will study different deep learning models, such as convolutional neural networks, recurrent neural networks, and autoencoders as well as probabilistic graphical models for deep learning like variational autoencoders. In addition, this course provides students with skills to implement and train deep learning models by exposing them to Python libraries such as TensorFlow. Finally, since some real-world applications require models to be interpretable, this course addresses the concept of explainable artificial intelligence (XAI).

Learning outcomes - Knowledge

By the end of the course, the student will

  • Know how to use and implement state-of-the-art deep learning models for different types of data.
  • Know basic techniques in XAI to explain models and their predictions.
  • Know how to use generative models to deal with limited amount of data.
Learning outcomes - Skills

By the end of the course, the student will

  • Implement and train state-of-the-art deep learning models in TensorFlow.
  • Explain the main difference between the different deep learning models and their learning algorithms.
  • Choose the appropriate methodology for a given learning problem and given data set.
General Competence

Students should be able to

  • Understand state-of-the-art literature on deep learning.
  • Re-implement deep learning methods for their own problems.
  • Provide interpretations for the predictions of deep learning models.
Course content

To address the learning outcomes listed above, the course will have the following content:

Generative modeling: Many datasets have limited training data or limited labeled data. The course introduces generative modeling techniques that can deal with this challenge by learning generative models that are able to create new observations. 

State-of-the-art deep learning architectures: Different data types and different input/output pairs require different architectures, e.g., a neural network architecture designed for time-series is not necessarily a good fit for tabular data. In this course, we will introduce and use state-of-the-art neural network architectures that fit different types of data, which will enable the student to make better choices when creating new models.

Explainable AI: Many real-world use cases require a model to be explainable. For example, the GDPR regulation requires that a decision made by a system should be understandable for a lay person. This part of the course will introduce some basic techniques for probing and understanding predictions made by a machine learning system.

Probabilistic Machine Learning:  Probabilistic modeling and deep learning have been succesfully coupled in methodologies such as the variational autoendocer (VAE). VAEs are not only useful as generative models, but also for making predictions on binomial or multinomial outcomes. Thus, VAE predictions are drawn from a probability density funcion, which allow us to say something about their degree of certainty.

Teaching and learning activities

Learning activities will combine lectures, case discussions, 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. 

Software tools
Software defined under the section "Teaching and learning activities".
Additional information

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.

 

Qualifications

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.

Required prerequisite knowledge

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
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: 
  • Bilingual dictionary
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: 
1 Semester(s)
Comment: 
Written assignment.
Exam code: 
GRA 41622
Grading scale: 
ECTS
Resit: 
Examination when next scheduled course
Type of Assessment: 
Ordinary examination
All exams must be passed to get a grade in this course.
Total weight: 
100
Student workload
ActivityDurationComment
Teaching
36 Hour(s)
Seminar groups
14 Hour(s)
Student's own work with learning resources
130 Hour(s)
Sum workload: 
180

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.