Beskjeder

Dear all, we hope all is well with final exams and project 3. We have been made aware of the fact that many of you this year have exams close to our deadline for project 3, December 15 (there was an error in devilry btw). 

We decided thus to move the final deadline to December 18 at the devilrish time 2359 (1159pm), hoping this can be of help. That is the final deadline.  It will be updated soon in devilry.

If you have not received feedback on project 1 let us know. Feedback on P2 is being added as of now and by the end of this week it should be there. It has taken us more time than expected to finalize the feedback. With 130 participants and fantastic projects we have clearly underestimated the time it takes to wrap this up. Also, by our bylaws on grading, every project has to be graded by two of us. We apologize for the delay in obtaining the final feedback. Anyway, the quality on all projects is very high and...

Dear All, this is or last week of organized teaching activities. Today is also our last lab session this semester and our last lecture is Thursday (no lecture on Friday).

We will during Thursday's lecture summarize what we have covered this semester with an eye on possible research projects in machine learning, advanced courses and other topics of interest. We will also point to other topics such as Bayesian statistics and Bayesian machine learning as a teaser for future research directions.

In the meantime, best wishes to you all and thx so much for all heroic efforts this semester. It has been a great pleasure to get to know you all,

Hanna, Lucas, Morten, Stian and ?yvind

Dear all,

here's as promised the link to John Aiken's lecture on XGBoots. I hope you may find it interesting, and again thanks to all of you who presented possible data sets last Friday, cool.

John's slides are at https://docs.google.com/presentation/d/1A5tMrZSa7XBwSZMEFDkkfmEBlDnM1y_H_DEhu5eWJDU/edit#slide=id.p

Else, this coming Thursday we will finalize our material on Support Vector Machines, both for classification and regression, with the final equations and code examples. 

Hastie et al's chapters 4.5 and 12.1-12.4 cover much of what we plan to discuss and have discussed. The slides at ...

Dear all, we hope all is fine with project 2. The final deadline as we wrote earlier is this Wednesday at 2359 on devilry.ifi.uio.no

Project 3 is now available at the github address of the course, scroll down https://compphysics.github.io/MachineLearning/doc/web/course.html and you will project 3. The deadline is December 15.  For this project you can propose your own data sets and we invite you all to send us a proposal that you can present during Friday's lecture this week. Your presentation should include a description of the data set with some motivation on why you think this data set is interesting and a brief plan on which methods you plan to explore. The presentation shouldn't last more than 5 min. Please send me your proposal by Thursday this week. We will also have a presentation of XGboost on Friday.

&nb...

Dear all, since many of you have asked for a small extension (and since we are delayed with the feedback, apologies from our side here), the regular deadline has been extended to Wednesday November 13 at midnight, that is 24 (or 12am). This is has been implemented in devilry as well (note that devilry has 2359 or 1159pm).

For those of you who have been sick or have other approved reasons the deadline is Friday November 15, at midnight (please upload however your github link to devilry by Wednesday).  

We hope this helps those of you who run into various problems this weekend.  The new deadline will not be changed however, else we end up with too many delays with project 3.  See also the weekly info to be sent out later today. Project 3 will be available this evening.

Best wishes,

Morten

Dear All, we hope all is well with project 2. The deadline is approaching and as we wrote in a previous post on Piazza, devilry does not allow us to set a soft deadline. It means that we would like to propose the following procedure:

1) upload to devilry.ifi.uio.no your GitHub/GitLab or similar link where you have your report, codes and data files. Make sure you have a Readme file which explains where we find the material. Structuring the repo in folders like Report, Codes etc also helps with respect to the readability.  Do this by the deadline Friday 8 at 23.59!!

2)  Then, in case things go wrong or you need some more time to brush up things, make sure you upload (Git gives a timestamp) your report and codes not later than Monday the 11th of November at noon (12pm) to your Github/Gitlab or similar repository. 

3) Fo...

Hi everybody,

we hope all is going fine with project 2. Last week we started discussing decision trees and the algorithms for setting up the trees such as the ID3 algorithm. We continue this week with the ID3 algorithm and the so-called CART (Classification and Regression Tree) algorithm. We will discuss these algorithms in more detail with code examples etc this Thursday before we move over to bagging, boosting, bootstrapping and random forests this Friday. This material is well covered by Geron Aurelien's text chapters 6 and 7 and lecture slides on decision trees (some minor updates to come later today). The text by hastie et al has a discussion of these topics although not very satisfactory) i chapters 9.2, 10.1-10.4 and partly 10.9.

Best wishes to you all and see you tomorrow.

Hanna, Lucas, Morten, Stian and ?yvind

Dear All, we hope this week started the best possible way for all of you!

Last week we started our discussion of dimensionality reduction methods like the principal component analysis. We continue with this this Thursday (see the lecture slides at https://compphysics.github.io/MachineLearning/doc/pub/DimRed/html/DimRed.html and also chapters 14.5-14.7 of Hastie et al, slides almost done)

We will try to conclude with the dimensionality reduction methods tomorrow in order to begin with our next algorithms on Friday, decision trees and random forests (see lecture slides at ...

Hi all, since many have asked and it is now only in a discussion thread, I share here ?yvind's GitHub link to his lectures this Thursday https://github.com/schoyen/fys-stk-lecture

We will continue partly with this on Thursday this coming week. On Thursday we will also discuss more about project 2 (credit card data) and technicalities around project 2.   

We will also discuss various preprocessing aspects (normalizing data etc) as well principal component analysis and other dimensionality reduction methods. These are all topics of interest when we look at more complicated data sets than say the simple Franke function we had in project 1.   

The slides  for this (not yet completely ready) can be found at https://compphysics.githu...

Dear all, we would like to wish you all the very best with the finalization of project 1. Concerning the deadline, devilry allows us as teachers only to have hard deadlines, that is the time you find on devilry.ifi.uio.no, which is today at 23.59. Now, in order to be sure please upload your GitHub link before the deadline. Your GitHub link for the project should contain the report, codes and eventual additional material used to test your codes and outputs.  In case you don't meet the deadline for uploading the report, this gives you some flexibility. GitHub has a time stamp which indicates when you uploaded the last version of the report.  In order to give you some flexibility, we allow corrections till Wednesday at 3pm. If you hand in later than that, we will, in order to be fair to those of you who handed in on time, have to deduct 10 points from the final score of 100 for every day you are delayed. 

This however does not apply if you...

Hello everybody and thx so much for heroic efforts with project 1.

I have now posted the first iteration of project 2 and we will discuss this during tomorrow's and Friday's lectures. A git pull should give you all relevant files, or simply go to https://compphysics.github.io/MachineLearning/doc/web/course.html and scroll down to project 2.

The main aim of project two is to develop a feed forward neural network code for both classification and regression. We will discuss the development of such a code during the lectures tomorrow and Friday. The code examples and more can be found in the neural network slides at https://compphysics.github.io/MachineLearning/doc/...

Hello folks, here comes a quick update with plans for this week and project 1. 

Last week we discussed logistic regressions and gradient descent (SGD) methods and we barely started scratching the surface of the stochastic gradient descent family of methods. This will be the topic for Thursday's lecture (with possible migration into Friday's lecture). We will discuss how to set up the SGD for both linear regression and logistic regression (classification) problems with examples. We will also discuss how to use automatic differentiation, an extremely useful algorithm included the autograd library. On Friday we move on to neural networks. That will keep us busy next week as well. 

The material for the latter can be looked up in for example the slides at ...

Hi all, this is the weekly fys-stk3155/4155 digest. Last week we started discussing gradient methods, a central element of all ML methods. We started at the end of Friday's lecture to discuss logistic regression as well. The main focus this week is indeed on logistic regression, theory and examples as well as a continued discussion on gradient methods. project 2 deals with a classification problem and we will develop code for both logistic regression as well as for neural networks. We begin with neural networks next week. 

This Thursday 

we continue our discussion on Logistic regression as well as a further discussion of gradient methods, see the slides at 

https://compphysics.github.io/MachineLearning/doc/pub/LogReg/html/._LogReg-bs000.html for logistic regression and ...

Hi all and welcome back to the weekly digest with plans for rest of the week.

First, we apologize for the confusion with the lab room this Tuesday. This Tuesday as well as Tuesday Oct 22 are the only days when we will back to VB3, else we are at F?434 for all remaining Tuesdays. We forgot to remind you about this. Please accept our apologies here. 

This Thursday we continue our discussion of resampling methods, with an emphasis on cross-validation, test and training errors, and the bootstrap method. The discussion will be...

Dear all, again, we hope this week started the best possible way. What follows is the weekly digest from last week with plans for this week.

Last week we continued our discussion of Linear regression with an emphasis on ordinary least squares.

We used last Thursday to remind ourselves about some basic elements of statistics and statistical analysis, such as quantities like the mean value, variance, standard deviation and the covariance. Parts of this material can be found under the slides on 'elements of probability theory' https://compphysics.github.io/MachineLearning/doc/pub/Statistics/html/Statistics.html and the beginning of the slides on Bayesian st...

Dear All, two small messages about exercises and projects and a  reminder:

1) Project 1 (first iteration, let me know if you spot typos etc) is now available, see for example https://github.com/CompPhysics/MachineLearning/tree/master/doc/Projects/2019

2) The second homework is also available at https://github.com/CompPhysics/MachineLearning/tree/master/doc/Projects/2019

Or, if you have cloned the whole repository, a git pull gives you the latest update.

Reminder: recall that the Thursday lectures ha...

Dear All, I hope the semester start has been a smooth one and that you have had an enjoyable summer. Welcome to FYS-STK3155/4155.

Our first lecture is Thursday Aug 20 and due to an unfortunate error by the study administration we were assigned a lecture hall with too small a capacity.

Unfortunately this means that our lecture starts at 1415 instead of 1215. I hope this does not cause too many problems and I apologize again for this unfortunate situation. It was impossible to swap auditoria right before semester start.    

It means that our lectures will be

Thursday 1415-16 in Store Fys Aud, Dept of Physics

Friday  1215-14 , same place.

The time for the lab is still Tuesdays 8-16. In case there is need for more lab time we may expand.

Else, all material is at the GitHub site of the course&n...

Dear All, the first lecture is Thursday 22, 1215pm in Store Fysiske Auditorium, Department of Physics.

Lectures are 

Thursdays 215pm-4pm

Fridays 1215pm-2pm

Lectures are in Store Fysiske Auditorium, Department of Physics (third floor, western wing of the building)

The first lecture is Thursday August 22. The last lecture is Friday November 29. The first lab session is Tuesday August 27.  

Depending on how many register for the course, we may divide the lab session into four groups (or open for more lab sessions) as follows:

Group 1: Tuesday 815am-10am

Group 2: Tuesday 10am-12pm

Group 3: Tuesday 12pm-2pm

Group 4: Tuesday 2pm-4pm

All course material is available at the github address https://github.com/CompPhysics/MachineLearning

More details about prac...