Messages
The admin contacted me to say that since you are doing the project in groups, you should submit it in Devilry, not Inspira, even though it's a Home Exam. Apparently Inspira can't handle groups.
You should be able to see your group's new deadline now on Devilry. If not, contact me.
I hope you did well in the exam. You can take a look at the solutions to get an idea of what a paper getting full marks would look like and get some sense of your possible grade. If you find something unclear in these solutions, or you think there is an error, please post a note on Piazza.
Best luck with the rest of the exams and your final project report!
Christos
Since for some reason the course page has not been officially updated (and I can't update it myself), this is the actual exam format:
Two projects (i.e. take home exams), for 40% of the grade
One written exam, for 20% of the grade.
The deadline for the second project will be the 13th of December. Until then, take advantage of intermediate deadlines and Piazza for questions.
If you have time, take a look at the new data in assignments together with a code for testing it. The train/test data is generated from the same distribution, while the validation data is generated from an imaginary recession.
You can watch two short videos, and we'll then do a flipped classroom next week. There will be a lecture next week.
Dirk will do an extra session on Thursday at 8 in Seminarrom Shell this week (17.10) since he has a scheduling conflict next week.
14:15-16:00
https://hangouts.google.com/call/IghDiXrANWltB8KWJAfxAEEI
There will be a guest lecture by Claire Emma Birnie, Senior Data Scientist at Equinor on the 10th, and a long session (8-12) on Oct. 17th. The room will be announce shortly.
A couple of notes for your Project 1 Implementation.
1. The name of your class should not literally be NameBanker, but 'your-team-name'Banker, e.g. ChristosBanker. This way I can test all of them later without name-clashes.
2. Make sure your code runs on TestLending.py, and that you do all your hyperparameter tuning in your .fit() function.
Project 1 is now available. I already alluded to how it is supposed to work. We now selected the deadline dates for 2019. The project is split into three parts. The first focuses on simple decision making and the other two on privacy and fairness respectively. You should submit something for every deadline to obtain feedback. The overall grade depends on your overall report quality, so you have a chance of improving upon your initial submissions as the semester progresses.
You can find the project in the assignments page, as well as here.
You will get a chance to work with the project in class on Thursday. Use piazza for any discussion related to the project.
Only use e-mail for personal questions such as sickness, etc.
The Jupyter notebooks from Dirk's sessions are available here. The repository will be updated as the lectures progress.
The university system for submission of assignments is Devilry:
If you missed the assignment because you didn't know this, you can email me.
Registration limits raised again for those wanting to take the course.
This assignment has two parts. The first is to write a small report and submit it to devilry by next Wednedsay. The second involves a discussion with the other students about your methodology, to be done in-class.
At the moment there are 44 people officially registered at the course, but there are only 29 registered on piazza. There is only room for 30 people in the labs, but we could squeeze in a few more. So, if you can't register but really want to, ask the administration if they can put you on a 'waiting list' and see whether those that are not attending can deregister.
23.8: Edited questions 3, 4, 12, to add some missing edge-case assumptions and question 12 as it was wrongly worded.
Please complete the introductory assignment individually (not in groups) by Wednesday next week (28.08). The assignment is not graded but it is used to gauge the level of background knowledge of the class. I will try and contrast it with the corresponding poll on Piazza :)
If Devilry is not available for you yet, simply email me.
All source code for the course and projects is available at
https://github.com/olethrosdc/ml-society-science/
The source is in the folder src/
References to the source code will be made in the slides whenever possible.
You are strongly encouraged to enroll on Piazza.
piazza.com/uio.no/fall2019/instk5000
Please also fill in the Prerequisites form in the Piazza course webpage. This will allow us to adjust our expectations and target the skills of the class as a whole more appropriately.
Note that the scheduling is somewhat irregular.
On odd weeks, we meet on for 2 hours on Wednesday afternoon and 2 hours Thursday morning.
On even weeks, we meet for 4 hours on Thursday only!