I will go over the exam from last year (2014) in tomorrow's group lecture(16.11 1415-1600). You should try to do it on your own first. Everyone in the course can come, regardless of group. This is the last week of lectures as well as group lectures.
Deadline for delivery is the 30th of october.
You will find everything needed in the folder "Mandatory Assignment 2" on the INF3490 semester site.
In the end of today´s lecture, I got a question about how the learn rate should be adjusted to the training set size, and if a larger set would require a smaller learning rate.
My answer was vague so afterwards with the help of my friend good Google, we found the following paper containing this table confirming that the learning rate should normally be smaller for a larger data set. Further, it confirms that updating weights after each training vector (online) is more effective than batch training (opposite of what our text book claims).
Unfortunately, there have been some problems with the Devilry access rights for the group teachers in INF4490. However, they should now be solved, and you will get feedback as soon as the group teachers have approved the assignments.
There will be no lecture Monday 28 September. However, the group teachers would be available in their slots during the week to help you with earlier exercises and answer questions you may have.
For those of you who aren't familiar with programming in Python, you can take a look at the lecture slides in the course INF3331, which gives a pretty detailed introduction to the language (the list of slides from 2013 can be found here). Alternatively, codeacademy has a short intro course here.
We have also added a short Python program (v2.7) that shows how you can read and use the data file in Python. You can find it here.
Deadline for delivery is the 25th of september.
You will find everything needed in the folder "Mandatory Assignment 1" on the INF3490 semester site.