Messages
A new exercise on unsupervised learning is available here. Check it out if you want some additional practice with this topic. This is a new exercise so please let us know on mattermost or through our email in3050-support if you have feedback or suggestions to improve it!
- Benedikte
We have now completed the grading of IN4050 and IN3050. The general impressions are that you have done a great job and solved the exercises to our satisfaction. Not at least given the difficult working conditions and the new form of exam.
Given the form of the exam and the availability of all possible means, we expected you to be able to get at least 60/100 points, and more than 95% of you did. Because of the difficulties this semester, we decided to lower the bar to 50/100. All the IN4050 students passed, and so did 170/174 IN3050 students.
We have published a proposal for solutions. Please take a look. We think that this detailed proposal, together with the explanation of the threshold for passing, more than answer the question of an explanation of the grade ("karakterbegrunnelse"), and we will not provide more detailed explanations for they who p...
I had a few questions on the perspective when interpreting directions ("right","left") in Question 8. The perspective is from above. In other words, going right in state 2 takes the cat to state 3. Going left in state 2 takes the cat to state 1. Which direction the cat is facing does not matter since the perspective is from above.
-Kai
Since I have got this question repeatedly
(3a) "Are we actually supposed to find the weights and the bias here? Or is it enough to give the expression where we dont know the weights and bias?"
Answer: You are supposed to find the weights and the bias
The target t=10 in (5b) is correct. It is not a typo.
We have received several questions through the official question portal: https://www.mn.uio.no/om/hms/koronavirus/brukerstotte-eksamen-v20.html
We handle these questions similarly as when we go a supporting round in the exam room in Silurveien ("tr?sterunde"). We answer the questions individually. Similarly to the "tr?sterunde", we do not publish the questions and answers here.
If there are mistakes in the exam set, or clarification we think all students should see, we will publish them here. We might also publish answers to questions we receive repeatedly to avoid having to repeat ourselves.
So far we have not seen any mistakes or serious obscureness in the exam set. In particular, we do not think there is a mistake in exercise (7a). You may stop asking about that one. Of course, there might be mistakes in other questions we h...
We have now published the trial exam. You find it here. We will publish solutions Wednesday 20 May in the afternoon.
Enjoy!
Kai, Jan Tore and the team
I have reproofread the slides for lectures 5-9 and found several errors that I have corrected. See here for overview of corrections made. The weekly pages now link to the corrected versions.
Jan Tore
Solutions for week 13 exercise can now be found here.
You can find the the Reinforcement Learning (RL) exercises here.
In the section
Implementation: how is PCA implemented?
PCA Algorithm
The text given in the function pca() states that pca_eigvec should have the shape (m eigenvectors, M dimensions). This is wrong, it's should be the other way around (M dimensions, m eigenvectors).
The assignment have been updated, and all the links to it on this site should now point to the new version.
This is the only change in this version and should not impact the code in any of the other sections.
The third mandatory exercise is out. You can find it here.
The deadline is April 30th, 23:59
The topic is unsupervised learning, with a focus on PCA and K-means.
The solutions to the week 10 exercises are also out.
Dear students,
We have some important updated information to you about the exam and mandatory exercises in the course.
First, the exam information is now updated on the course website under "Examination: Time and place". Please consult that page for the up-to-date information.
Second, due to the delays caused by the closing of the university, we cannot set aside as much time as we wanted for the third and final mandatory exercise, but still think it is important to give you practice with a wide range of the course topics - and have therefore chosen to give a shorter final mandatory exercise. The exercise will be focusing on unsupervised learning (the topic of the lecture on April 14), and will be published on April 14 with a submission deadline of April 30. When published, you will find it on the course website, under "assignments and deadlines".
Hope you are all doing reasonably well despite the extraordinary circumstances -...
This weeks's lecture is now out. Jeremy Barnes on Deep Learning. Enjoy!
An extra session on Neural Networks and Backpropagation was given by Marius, Friday 27 March. Afterwards, he recorded the main parts in two videos. They are published together with a Jupyter notebook here.
The extra session on neural nets and backpropagation will be given by Marius, Friday 27 March at 4 pm (16:00) in the Zoom room
Join Zoom Meeting https://uio.zoom.us/j/838582202
Meeting ID: 838 582 202
It was impossible to find a time slot that fits much more than half of them who replied. This was the most popular slot.
Thank you to everyone who took the time to fill out the poll on group session organisation -- and if you haven't, there is still time to have your say here.
Considering your comments so far, and the deadline of the second mandatory assignment, it seems you would welcome a recap session on neural nets and backpropagation. Therefore, we will be hosting an extra group session (via Zoom) this Friday, 27 March, to summarise the topic and answer your most frequently asked questions.
If you're interested in attending, please let us know which timeslots would work for you by filling out this poll. We will try to find a time that suits the greatest number of students, so please try to be flexible.
This week, week 9, 's lecture is out with slides, videos and some jupyter notebooks.
Dear IN3050/IN4050 students
This is an update on the situation for our course
We will resume teaching from Monday 23 March in digital form. We have chosen the following solutions. We have moved the program one week and will reduce the lectures towards the end of the semester, follow the schedule page.
1. Lectures
These will be recorded and uploaded to the course semester page before each Tuesday 10 am. We will use the weekly pages actively and distribute videos, slides, and other relevant material there.
2. Group sessions
There will be group sessions at the regular time slots, see
/studier/emner/matnat/ifi/IN3050/v20/timeplan/index.html.
These will be given interactively, using the Zoom tool, see...
The working period is prolonged, the deadline is postponed until 1 April.