Beskjeder

Hi all,
I wanted to remind you of our plans for labs till the deadline for project 3.
Next week there are no labs (also because many of the other teaching assistants are busy with own exams)
but we have two extra lab sessions, both at F?434
1) December 8, digital lab 10-12 and in person 12-14 and later as well if needed . 
2) December 15, digital 10-12 and in person 12-14 and possibly longer.

If you wish to brush up your skills about Tensorflow and Keras, the lectures from weeks 41-43 contain several examples. Also, the video here by two former teaching assistants (Kristian Wold, now PhD student at OsloMet working on quantum machine learning and Per-Dimitri S?nderland, now at Accenture)  can be useful, see
/studier/emner/matnat/fys/FYS4411/v21/forelesningsvideoer/ExerciseSessionJan26.mp4?vrtx=view-as-webpage

Never hesitate to send us emails!!  And we can always schedule either in-person mee...

Dear all, we plan to organize two more lab sessions prior to the deadline of project 3. These are

1. Wednesday December 8, 10-12 digital lab, 12-14 in person at F?434
2. Wednesday December 15, 10-12 digital lab, 12-14 in person at F?434

You are all most welcome and never hesitate to send us an email in case you wish to meet or set up a zoom session.

Best wishes,

Morten et al

Dear all, first thx a million for heroic efforts with project 2 and for all the exciting contributions to project 3 last Friday. 
This week is our last week of lectures and lab. The lab tomorrow runs as usual, with both in person 8-18 and digital labs 8-10 and 14-16 and the topic is obviously project 3.
However, since the deadline for project 3 is December 17, we will offer both a digital (10-12) and in person labs 12-14 on Wednesday December 8 and December 15 (there are no activities next week but feel free to send us an email). 
We are also planning to have the feedback for project 2 ready by December 10 or 11. 

Else, the topics this week are, besides our lab sessions on Wednesday,

 
Thursday: Support Vector Machines, classification and regression. 
Friday: Support Vector Machines and Summary of Course

Reading recommendations:
 o Geron's chapter 5.
 o Hastie et al...

Workshop plan Friday November 19
1215-1225pm: Are Frode Helvig Kvanum, Gard H?ivang, and David Andreas Bordvik, Next-day forecasts on spot prices for electricity
1225-1235pm: Lidia Luque, Voxel-wise multi-label brain tumor classification
1235-1245pm: Marcus Berget et al, Locating suspicious brain activity using neural networks
1245-1255pm: William Ho and Tom-Ruben Traavik Kvalvaag, Comparing semi-supervised learning and supervised learning for image classification

We will use the second part of the lecture for further discussions of projects 2 and 3 and a summary on boosting methods from last week and perhaps get started with support vector machines (our last topic). Feel free to bring your laptops.

We have plenty of time for questions and also possibilities to have additional presentations.
You are all welcome and this is an excellent o...

We have at present only four groups which wish to present possible data sets and variants of project 3 tomorrow. Feel free to suggest topics, these workshops are very useful and provide interesting inputs to the course. Please send me your suggestions by the end of the day today.  There is no need to make fancy and polished presentations, the important thing is to share ideas. Else, due to a conflicting workshop at OsloMet,  Thursday's lecture is cancelled.
Best wishes to you all with wrapping up project 2.
Morten

Dear All,
we hope this weekend has played out the best possible way for you all. Here follows an update with plans for this coming week. Please excuse us for  the length of this letter.

1) As you may have noticed, we extended the deadline for project 2 till November 20 at midnight. Note that we, by error, created random groups. This has been changed, please ignore that.
2) Project 3 is available and its deadline is set to December 17 at midnight. See https://compphysics.github.io/MachineLearning/doc/web/course.html and scroll down to project 3.
We have added extra scores for this project, meaning that you can get an additional 30 points if you do the optional exercise.
Note well that besides normal reasons for extensions (illness or similar valid reasons), we cannot make extensions beyond this data. If you hand in later than these deadlines, we will have to subtract points. 
3) For project 3, we would like to propose that we dedi...

Dear All,

last week we wrapped up the discussion on the few unsupervised learning methods discussed in this course, namely principal component analysis and clustering (k-means). We started discussing decision trees. We will this week discuss in detail the algorithms for regression and classification for decision tree and then move on to ensembles of trees, either via bagging (bootstrap with aggregate) , random forests, voting and boosting methods. 

The plans for this week are 

Wednesday: lab, in person and digital (8-10 and 14-16). Note that we will have pizza at F?434 around 1pm tomorrow, feel free to get some extra energy while wrapping up project 2. At the lab we will also discuss formats for the report.

Thursday Lecture: Basics of Decision Trees, Bagging and Voting

Friday Lecture: More on Bagging, Voting, Random Forests and start Boosting

...

Dear all, here's the weekly FYS-STK3155/4155 digest.

Last week we ended our discussions on deep learning methods, with an emphasis on convolutional neural networks and recurrent neural networks. We started discussing the principal component analysis method for dimensionality reductions. We will wrap up this part this week as well as discussing briefly another dimensionality reduction method like clustering and k-means. 

We will also discuss various aspect of project 2 as well at the beginning of Thursday's lecture.

After this, we start with our second last topic for this semester (most likely on Friday), namely decision trees, random forests and other ensemble methods (bagging, voting and boosting). 

This will keep us busy till the end of next week. Thereafter we move to the very final topic, support vector machines. 

We would also like to discuss possible data sets for project 3&nb...

Dear all,
here comes the weekly update for FYS-STK3155/4155.

Lab this week: We continue with project 2.  We would also like to remind you that we have (for those of you who cannot physically) digital labs at 815-10 and 1415-16. Feel free to use this also if you are attending the in-person labs. Today for example, for the lab from 815-10 in F?397 (which is a self-study lab), this can serve as an alternative. 

Last week we discussed convolutional neural networks. This week we will summarize our discussions of CNNs and discuss so-called Recurrent Neural Networks (of great interest for studies of tie series, and language sequences of variable length). As reading background , we recommend Goodfellow et al in their chapter 10. We also recommend Geron's chapter 14 and the lecture notes (some parts will be wrapped up later today).
We recommend also taking a look at Goodfellow et al and their chapters 11 and 12. These chapters contain many p...

Dear all, welcome back again.

Last week we discussed how to develop a feed forward neural network code. We introduced also how to set up a neural network using tensorflow and keras.  Many of these topics are described extensively in the textbook by Goodfellow et al, in particular chapter 6-12. These chapters provide a good overview of the topics we cover in weeks 40-43. This week, besides the introduction to convolutional neural networks, we will also discuss how to solve differential equations using neural networks.  On Thursday, we will summarize our discussions on feed forward neural networks and apply the theory to the solution of differential equations. This will be the topic for parts of Thursday's lecture. Thereafter we move on to discuss convolutional neural networks (CNNs) for the rest of this week. CNNs are well covered by Goodfellow's text chapter 9.

In summary, this week our plans are

• Lab Wednesday: Wor...

Dear all. Thanks a million for heroic efforts with project 1. 

Project 2 is now available and the first part builds on project 1 (it is a continuation of project 1) where you are going to replace your matrix inversion algorithm with a gradient descent algorithm for OLS and Ridge regression. Thereafter you will have to develop a neural network code for both regression and classification. 

At the lab Wednesday we recommend thus to get started with implementing this. Last week we discussed various gradient methods as well as developing the back propagation algorithm. 

This week we will discuss this algorithm and how to write our own feed forward neural network. This will serve as a useful starting point for project 2. 

The topics this week are thus:

• Lab Wednesday: Work on project 2
• Lecture Thursday: Deep learning and Neural Networks
• Lecture Friday: Convolutional...

Project 2 is now available at  for example https://compphysics.github.io/MachineLearning/doc/Projects/2021/Project2/html/Project2-bs.html

or go to https://github.com/CompPhysics/MachineLearning/tree/master/doc/Projects/2021/Project2 and fetch your preferred format.

Dear all and welcome to a new week!   
Here's a brief summary from last week with plans for this.

Lab on Wednesday: We still have the digital labs from 8-10 and 14-16, the zoom link is the same, 
https://msu.zoom.us/j/95317649875?pwd=aWM1akppam4yWVBIY29KaXE5cHpSZz09
Meeting ID: 953 1764 9875
Passcode: 536396
And the deadline for project 1 is Monday  October 11. It suffices to upload the link to your GitHub or Gitlab repository by the deadline. 
Project 2 will be available this coming Sunday and deals with neural networks.
This leads to this week's topics with a brief mention of what we discussed last week. Last week we summarized our discussion of Logistic regression for binary classification problems and started discussing optimization methods like various gradient descent methods. We will finalize this discussion this Thursday by studying the various stochastic gradient descent approaches that exit. The...

Dear all, we hope this week as well has started the best possible way.
Here follows the weekly digest, but first a change to the deadline for project 1.
We have changed the deadline to Monday October 11 at midnight.  It suffices btw to upload your GitHub or GitLab link to the where you have your report and codes. In the repository please also add a README file so we can easily find the relevant files. It is normally convenient to have for example three folders, one for codes you have developed, one for the report itself and perhaps another folder for different test runs (selected examples) which we can reproduce when we run your codes.
The report can be a doc-file (office or openoffice), a PDF file or a jupyter-notebook.

Else, this week we will discuss on Thursday we will wrap up our discussion on logistic regression from last week (with selected examples). Thereafter  we discuss how we can use various gradient methods to find the op...

Hi all FYS-STK folks, we hope all is well and that you got well started with project 1.
This week at the lab we keep working on project 1. We will keep the two digital labs as well as long as there is an interest. 

Last week we discussed resampling methods like Cross-validation, Bootstrap and jackknife, as well as discussing possible ways to understand what Ridge and Lasso regression mean. The slides from last week at https://compphysics.github.io/MachineLearning/doc/web/course.html (scroll down to week 37) cover what was discussed during the lectures. The link to the videos from the lectures can also be found in the weekly slides, as well as at lecture info on the official UiO website and the schedule link for the jupyter-book at 
https://compphysics.github.io/MachineLearning/doc/LectureNotes/_build/html/schedule.html

This week, the plan is to summarize our discussion on linear regression and what we have done these first weeks and...

Welcome back to a new and exciting ML week.

Last week we discussed Ridge and Lasso regression using linear algebra and the singular value decomposition of a matrix as well as linking everything with a statistical interpretation. There we noted that, using the maximum likelihood estimation ansatz, we could derive the ordinary least square equations. Using Bayes' theorem we were able to get an alternative way of deriving the Ridge and Lasso equations. 

This week we discuss resampling techniques and in particular we will focus on two widely methods used to obtain a better estimation of expected values, the so-called Bootstrap method and the cross-validation method.  We will also include a discussion of the bias-variance tradeoff and other statistical quantities like the estimation of confidence intervals. These are all topics of relevance for project 1.

By the end of this week you should thus have all theoreti...

Project 1 is available at https://compphysics.github.io/MachineLearning/doc/web/course.html, scroll down to project 1. The deadline is October 4 at midnight.

You can also retrieve all files at (html. pdf, tex, ipynb) at https://github.com/CompPhysics/MachineLearning/tree/master/doc/Projects/2021/Project1.

If you find typos or unclear issues, please let us know asap.

It will also be discussed during the lectures in the coming weeks.

Dear all, we hope this week has started the best possible way for you all.

Here's a short overview of the plans for this week with a recap from last week. 

Last week we discussed ordinary least squares (OLS), Ridge and Lasso regression in terms of various linear algebra features, in particular w emphasized the role of the singular value decomposition and how we could interpret the regularization terms in Ridge and lasso regression. This material is available from the weekly slides of week 35, see the general overview of the weekly material at https://compphysics.github.io/MachineLearning/doc/web/course.html. At the end of the slides you will also find the exercises for this week. These exercises are meant as stepping stones towards project 1. The latter will be presented on Friday this week.

This week we plan to wrap up our linear algebra discussion of the regression methods (with some simple examples) and move on to a statistical...

Dear all, first a great thank you for having chosen FYS-STK3155/4155.
What follows is a lightly longer mail, with weekly updates and information about the computational labs.
Plus other information.

Concerning the exercises and projects, for those of you searching for lab partners, please fill in your information at https://docs.google.com/forms/d/12VNXJOqMfLGism580eBps_M7zk-gzXe7Qd-B2Ll_s8o/edit
Based on your responses we will by the end of this week come back with group proposals.

Else, for our computational labs and exercise and project sessions, we have at our disposal rooms F?397 and F?434 at the Physics Building (eastern wing) from 8am till 6pm every Wednesdays (two small exceptions from 2pm in September for F?434).

The schedule is as follows

Room F?397
--------------
8-10am: Selfstudy, the lab is open but no lab group. Feel free to come if you wish.
10-12pm: Computational Lab group 3...

NORA AI competition: See the link here https://www.nora.ai/Competition/image-segmentation.html

NORA AI competition: See the link here https://www.nora.ai/Competition/image-segmentation.html

For the lab session on Wednesday August 25, the following link to Python programming, Git, Github/Gitlab and more, may be useful.

We will cover some of the material here this coming lab session. 

Dear all, 

first a warm welcome to all of you. We hope you will enjoy the course.  Our first regular lecture is Thursday August 26 and you find the link to the educational material at the  GitHub link here, and for more easy display of the weekly lecture slides, we recommend the weekly overview link.  

Our first lab session is Wednesday August 25 and the aim is to present to those of you who need a repetition of Python and programming some basic libraries we will use in this course. Each lab group last two hours. The first hour will be dedicated to Python libraries and how to install them and a quick reminder on programming. The second hour will be dedicated to version control software like Git and how to use repositories like GitHub/Gitlab etc.

The lab will ta...