Semester page for FYS4411 - Spring 2021

Dear All,
we hope you are doing find in these odd times. Unfortunately it seems that we may most likely end this semester as well with an online version of the course.  Too bad. 
However, we keep our group discussion structure throughout the semester. 
This coming Thursday is our second last regular session. Our final session is Thursday May 6, next week.  After, we can always organize own discussion meetings, but our last regular session is May 6 from 1415 to 19.
We begin this Thursday at 1415 with our groups:
Group 1 and breakout room 1: Time-dependent Hartree-Fock theory. ?yvind will continue discussing the theory during the first 1-2 hours, thereafter we work on the project.
Group 2 and breakout room2: Machine Learning and many-body Physics: Morten will discuss the link between Boltzmann-Machines and Neural Networks the first 1-2 hours. For this part, Bendik Samseth's master of science thesis is an excellent read, see http...

Hi folks, we are hoping all is well. Just a very short message about this coming Thursday.
Since ?yvind and myself have to attend two different meetings before 4pm, we propose that this Thursday we begin at 4.15pm/16.15 instead of our regular time. 
Next week we start again at 2.15pm/14.15.
We apologize for this change. We hope however we can discuss properly the various projects.
Best wishes to you all,
Morten and ?yvind

Dear all, after last week, we have now three project groups for FYS4411/9411. There is a small possibility of having an additional one on fermion VMC, but our three basic working groups are

  o Time-dependent Hartree-Fock with material at https://github.com/Schoyen/tdhf-project-fys4411/tree/master
  o Neural Networks and Boltzmann Machines with material at http://compphysics.github.io/ComputationalPhysics2/doc/LectureNotes/_build/html/boltzmannmachines.html
  o Quantum Machine Learning with material at http://compphysics.github.io/ComputationalPhysics2/doc/LectureNotes/_build/html/basicquantumcomputing.html (this material is in the making.

The jupyter-book at https://compphysics.github.io/ComputationalPhysics2/doc/LectureNotes/_build/html/intro.html contains much more material, also on Hartree-Fock theory and other many-body methods. The TDHF and the QML projects will be updated tomorrow with more text.

Our Sc...

Dear all, welcome back to FYS4411/9411 after the break. 
Since we are going to start with project 2 tomorrow, we will use the first hour of the lab session to present the projects. There is no lecture tomorrow since we will use the lab session tomorrow to figure out how to organize the rest of the semester. 
We have at present four projects:
1) Variational Monte Carlo for Fermions, continuation of project 1
2) Machine Learning approach to project 1, continuation of project 1, emphasis on Boltzmann machines
3) Time-dependent Hartree-Fock theory
4) Quantum Machine Learning.

We start at 1615 tomorrow. Some of you have also an extension for project 1 and after an introduction to the new projects, we will split into working groups and plan the rest of the semester depending on your interests. 
We may have to plan separate lectures/sessions for each group.

The most important thing, no lecture tomorrow and we ...

Hi all,  a brief message here:
This is our last lecture before the break, and the last one concerning project 1.
Last week we discussed how to use and install MPI, this week we discuss in more detail OpenMP and how to implement that in project 1. We will also discuss other issues around project 1. Feel free to send us an email before the lecture and we may hopefully be able to bake them in. We may end a little bit before 4pm and will then just start with the lab. 
Else, as written earlier, we are pretty flexible with the deadline. We would however like that you upload  the link to your github address by the deadline. 
Also,  feel free to come with suggestions about topics for project 2.

We would like to have some feedback on which of the following topics is the more interesting one:

-Fermion VMC, continuation of project 1
-Deep learning applied to project 1, either neural networks or Boltzmann machines...

Dear All, we hope you are doing well during these difficult times.  Here's a brief update with plans for this and next week.
As you may also know, we are in the unfortunate situation that we cannot have in-person lectures and in-person labs.  This situation is going to persist till after the holiday break. If we are unlucky, it is not unlikely that the rest of the semester will have to be fully digital. 
Next Thursday it is our last lecture before the holiday break. Our next lecture is April 8 and we will have our last lecture May 6. The last lab session is May  15.

Feel thus free to schedule eventual zoom sessions if needed in addition to our Thursday sessions. 
Hopefully we can then compensate for the lack of in-person contacts.
Due to the present situation and the difficulties in getting things done, we are going to be pretty flexible with the project deadlines. We have set a given deadline for your planning of the s...

Hi all, we hope you are doing well. Here comes the standard weekly update with our plans for this and next week.
Last week we attempted at a conclusion on gradient methods and for project 1, the simplest plain gradient descent does a pretty decent job. 
What we have now is
1) a VMC program with brute force and importance sampling
2) hopefully added a gradient descent optimization in order to get the best variational parameter(s).

Now we need to think more carefully about statistical errors and we are going to encounter two much used techniques, the so-called bootstrap technique and the blocking method. 
Hopefully these methods (to be discussed this week and next week) allow us to get a better estimate of the standard deviation and thus our statistical error. 
The learning material is at for example http://compphysics.github.io/ComputationalPhysics2/doc/pub/week8/html/week8-reveal.html or as a jupyter-notebook at http://...

Dear All, we hope you are all doing well. This is the weekly FYS4411/9411 update with plans for this week and links to various topics. 
Concerning the covid-19 situation, as of now (today, Wednesday Feb 17), we do not know if we can have in-person lectures and labs. Hopefully, with an improved situation this may change to the better.

Else, last week we attempted on giving you a link to the derivations of the Fokker-Planck equation and the Langevin equation used in importance sampling. For us, the most important aspect is however how this is implemented in our codes. 
We ended last week with a start on optimization methods. This is were we start this week and we will continue with this topic next week as well before we move on to resampling methods and finally parallelization. 
These will be the last ingredients we need in order to have a professional variational MC code.

The lecture material this week is at for example htt...

Hi all, we hope this week has evolved in the best possible way. Here's out weekly digest with plans for today's lecture and lab.
Last week we started discussing the practicalities on importance sampling. Our focus was on the elements we need to include in our program. Today we will wrap up our discussions by looking into the theory of Green's functions, the diffusion equation and the Fokker-Plank and Langevin equations. 
Towards the end of the lecture we start our discussion of gradient methods. This will be our next programming element and is needed in order to be able to find the optimal variational parameters before we launch a large VMC calculation. The slides are partly those from last week (Feb 1-5) and on gradient methods, the slides are at
for example http://compphysics.github.io/ComputationalPhysics2/doc/pub/week5/html/week5-reveal.html, or just scroll down the page http://compphysics.github.io/ComputationalPhysics2/doc/web/course and you will fin...

Dear all, welcome back to FYS4411/9411. Here follows a quick review of what we did last week, with plans for this week and reading material.
Again, due to the present covid-19 situation, the lecture and the lab sessions are still online. We hope to be able to meet in person as soon as possible.

Last week we discussed the Metropolis algorithm with and without
importance sampling and derived the Metropolis-Hastings algorithm.

We replaced the brute force Metropolis algorithm with a walk in
coordinate space biased by the trial wave function. This approach is
based on the Fokker-Planck equation and the Langevin equation for
generating a trajectory in coordinate space. The link between the
Fokker-Planck equation and the Langevin equations and
importance sampling is non-trivial and perhaps almost impossible to
explain without a full semester course in for example statistical mechanics
and transport theory. However, we ...

Last week we discussed the basic setup of a Variational Monte Carlo code for solving project 1. In setting up a C++ code for variational Monte Carlo calculations, we will use an excellent framework developed by a former Computational Physics student, Morten Ledum, now PhD student at the Hylleraas center for Quantum Chemistry. The GitHub repository is at https://github.com/mortele/variational-monte-carlo-fys4411.

At the lab today we will continue our discussions on how to build up the code while during our lectures we will discuss in more detail the Metropolis Algorithm, Markov Chains and Importance sampling.
These topics will be continued next week as well.
These technicalities are discussed in the lecture slides for this week, see the slides at http://compphysics.github.io/ComputationalPhysics2/doc/web/course and scroll down to week 4. 
In particular we will start discussing Importance sampling. 
Recordings of the lectures from the t...

Due to the present covid-19 situation in the municipality of Oslo and present restrictions, we have to stay with digital labs and lectures another two weeks till new information is available.
Note that we have a made a change of lecture hall (Lille fys auditorium ) and lab session room (F?434). These rooms have larger capacities and as soon as some of the restrictions are loosened, we may hopefully meet in person.

Last week we startet with variational Monte Carlo methods and looked at specific equations we will need in order to work with project 1. 
This week we will go in more details on how to build a VMC code as well as discussing the Metropolis algorithm and Markov Chains (a topic not all of you are familiar with). 
In our discussions on how to build an efficient VMC program, we will use as an example an excellent layout by a former Comp Physics student, now PhD student in computational quantum chemistry,  Morten Ledum....

Dear all, since it is not clear how long the restrictions will last, we publish here the zoom link to the lectures:

Topic: FYS4411 Computational Physics 2
Time: This is a recurring meeting Meet anytime

Join Zoom Meeting
https://msu.zoom.us/j/6424997467?pwd=L28xL3NqVno5MU5OUVdKUExYdnA5Zz09

Meeting ID: 642 499 7467
Passcode: CompPhys2
 

Irrespective of whether we have physical or online lectures, all lectures will be recorded. Welcome to you all.

Dear all, we hope you have had an excellent winter break and are ready for gaining new and exciting knowledge.

Our first lecture is Thursday January 14, at 215pm. We will have weekly lectures (except for spring break in week 13) 215pm-4pm every Thursday till May 6.

Depending on the capacity of the room, we may run these lectures in person. These lectures area also available via zoom.

The lectures are followed by an in-person and digital lab. The lab starts at 415pm and ends at 7pm.  The first lab is January 14 and the last lab is May 6. 

All teaching material is available through the github address of the course, see https://github.com/CompPhysics/ComputationalPhysics2

Welcome to a new semester.

Morten and ?yvind.