Messages
Dear IN3050/4050 students,
We are done grading, and your grade should arrive shortly. To give you a little context to your grade, we have put the grades distribution for 2021 here.
Thank you for the good work this semester - have a nice summer!
Best wishes,
Jan Tore and Kai
Dear students,
We have posted new versions of the exam set and their solutions on the "exams" page on the left. The updates are relatively small and do not really change the final answer on any problems. So, if you have already looked through the solutions, there is no need to do so again. However, if you haven't yet, please look at the new, updated version.
-Jan Tore and Kai
Hi everyone!
The exam set and suggested solutions are now online here. Thanks for your work and attention this semester, hope you have a chance to enjoy the warm summer weather before (or while) preparing for any further exams!
Have a nice summer!
-Jan Tore and Kai
Hi everyone!
Here we will post clarifications/corrections to the exam text.
Problem 2 - Fitness Sharing: There seems to have gotten in 2 small errors here. 1) The value alpha (=1) does not belong in the formula. You can ignore that value. 2) The problem refers to the fitness value you calculated in a. That is incorrect, what is meant is the fitness value that is listed in the table on the left. In a, you actually calculated selection probabilities, not fitness values.
Problem 3: " 'dollar' occurs. number" is the number of occurrences of the word 'dollar' in the document.
Problem 5: The learning rate is given in (5a).
Problem 7b: If you're wondering what policy the agent used here, the exact policy doesn't really matter - but the actions chosen by its policy are the...
Hi everyone!
The suggested solutions to the 2020 exam, which we cover in last week's and this week's lecture, are available here.
-Kai
Hi everyone!
We have received several questions regarding sources and citations. The exam in IN3050 and IN4050 2021 will be similar to an exam in Silurveien. Any of the questions asked could have been asked if the exam had been given in Silurveien. If you have learned your syllabus, the questions can be answered in the usual ways without accessing books, the internet, or other sources. If you answer as usual, you do not have to cite your sources. In particular, if you know the answer to a question, you should not waste time finding out where in the syllabus the knowledge you apply can be found.
If, however, you look up something in the syllabus and you quote it directly, you should give a simplified citation of where you found it, e.g., "Marsland, p. 999" or "Lecture 30, slide 99".
If you make active use of any sources outside the syllabus, you should make proper citations.
Hi everyone,
Since I didn't finish large parts of the EA, RL and PSO problems from last year's exam, and since it was requested in today's lecture, I will continue going through last year's exam next Wednesday. We may not use the full 2 hours, but if you're interested, join us to see the parts of the exam that I didn't get through today.
See you!
-Kai
Hi everyone!
We have assembled the information about this year's exam format here. The information will be repeated on the exam, but we recommend that you have a look now to know what to expect beforehand.
-Jan Tore and Kai
Some students have asked for a list of the syllabus. During the semester we have listed this in the weekly pages as mandatory readings. We have now in addition gathered all the mandatory readings at the syllabus page.
Hi students!
In the next interactive session, we will go through the suggested solutions to last year's exam. We recommend that you have a look at it in advance, it is available here.
-Kai and Jan Tore
Hi everyone!
We made a small correction to the trial exam, as we noticed the perceptron-task was missing a learning rate. The correct rate should be 0.1. This is now updated in the online PDF.
-Kai
Dear IN3050/4050-students,
We have prepared a trial exam for you that will give you a good idea about the types of tasks and the size you can expect on the final exam. On the final exam, bachelor students and master students will have slightly different versions, with an emphasis on on a bit deeper understanding/reasoning about algorithms in the master-student exclusive questions. Here, we have combined the master student and bachelor student questions in the same file, so you can see both. Note that this makes the trial exam appear a bit longer than it really is: On the final exam, you will only see the questions relevant for your level.
Wednesday next week we will publish the suggested solutions and we will go through them together in the interactive session. We encourage you to take the time to try to solve all the problems yourse...
Hi everyone!
Since the exercises in the course are now all done, group sessions are also over from today. There is one exception, which is Marius' group at Mondays 10:00 (group 3), which will continue, reviewing some central parts of the course syllabus. We may also add other groups as exam preparation begins, but it depends a bit on the desire of you students, as we will also devote 2 interactive sessions to reviewing the trial exam and last year's exam.
-Kai
I have finally filled in the readings for week 10. No mandatory readings, but I have given some recommendations if you are interested in reading more.
Jan Tore
There will be no group sessions in IN3050-IN4050 for the period between Mar. 25 and Apr. 6, our virtual week 11.. We will distribute Mandatory 3 before Easter.
To get one-on-one help from a group teacher, go to the helpline room in our mattermost channel and ask for help. A group teacher will contact you (as time permits, first come first serve). Be prepared for the chat, have questions and code ready.
All groups are now digital. There is no booking system by now. You are free to switch group. Some complained during today's interactive session that the Friday group 6 is too crowded. You are welcome to change group.
The university's buildings will be closed to students from noon today (16/03-21). All groups will from now on be digital. See more information here: /english/about/hse/corona/news/reading-rooms-and-libraries-will-close-to-students.html
There was a typo in mandatory 2, question 2.4. You should here train the multi-layer network you made in exercises 2.3 on the (X, t2) data. The network for the (X, t) dataset was already considered in question 2.3.
The notebook from Marius, Wednesday Mar.10, is now out. You find it here.
The main part of the interactive session March 10 will be Marius Havgar guiding us through a concrete example of the backpropagation algorithm. Zoom link.
Jan Tore