setup and solve typical machine learning problems, by implementation or by using established computer simulation tools.
decide which machine learning methods/algorithms are suitable for which type of learning problems, i.e. know about their most important weaknesses and advantages.
decide how to represent data to facilitate learning.
recognise typical effects of bad initialisation and parameter selection and suggest ways to improve the results.
describe how, and why, machine learning and natural computation methods work, explain principles and show examples.
The course introduces various machine learning techniques, with a focus on natural computation methods. The course is divided into a theoretical part (4 credits) and a practical part (6 credits).
The theoretical part consists of lectures and written course material on various topics, including (but not limited to)
Learning paradigms (supervised, unsupervised and reinforcement learning). Artificial neural networks for classification, function approximation and clustering. Reinforcement learning methods and Temporal Difference Learning. Evolutionary computing (genetic algorithms, genetic programming). Swarm Intelligence (ant colony optimisation, particle swarm optimisation).
The practical part consists of lab assignments and a project assignment. The subject of the project assignment is open for the students to define themselves, but must be approved before the work begins.
Lectures, lab assignments and a self-defined project assignment.
Written or oral exam, and through written reports covering the lab assignments and the project assignment. The students are also required to present their results from the project assignment in seminar form.
This course cannot be included in the same degree as 1DT022 or 1DT646.