The computer assignment consists of implementing one method for input selection.
The methods are tested on the test problems given in this page.
The topics of the assignments are decided individually. For questions please
contact the course assistant.
Reporting is done using the given latex or word template. Deadline for returning
the reports is 22.12. The report can be returned by email in PDF format. The students
should report the relevant results and also the method they implemented in a general level.
Each student tests one method.
Currently public matlab code related to the project is available
The final goal is to have a matlab toolbox with a graphical interface.
The test problems are available
See also the slides of the first lecture.
The electricity consumption time series should be divided into windows of length 20 to generate the inputs.
Generate the scalar output corresponding to direct 10 step ahead prediction.
On the boston housing data choose 380 data randomly
from the whole set for training and use the rest to calculate the test error.
It is a good idea to normalize data before input selection and regression!