Messages
Dear all, yesterday was our last regular session, with an emphasis on project 2 only. For those of you interested in using NetKet and their neural network implementations, Daniel sent me an excellent link for their manual. The link is https://www.zora.uzh.ch/id/eprint/220384/1/ZORA_pdf.pdf
If you wish to, you can compare your own restricted Boltzmann machine with their implementation, or if you wish, venture into using their neural network implementation.
Else, feel free to shoot us an email.
Best wishes for the the rest of semester to you all,
Morten and Øyvind
This week we will, after a brief summary of what we have done, focus on project work only. This is our last session as well.
Dear all, we hope the week has started the best possible way for you all.
This week we will have our last lecture on machine learning approaches related to one of the variants of project 2. We will summarize the essential equations needed for generative models like Boltzmann machines and then move over to how we can simulate the same system using neural networks. We will also try to elucidate the differences and point to codes for setting up neural networks.
Next week we will, after a brief summary, focus on discussing and working on the projects.
See you soon and best wishes to you all,
Morten and Øyvind
Dear all, welcome back to a new exciting week with FYS4411/9411. This week we will go through the nitty gritty details (mathematical ones) for setting up a so-called deep generative neural network using so-called Boltzmann machines. The slides in the form of a jupyter-notebook at https://github.com/CompPhysics/ComputationalPhysics2/blob/gh-pages/doc/pub/week14/ipynb/week14.ipynb will form the basis of the lecture. The lecture will mainly be a classical whiteboard one (almost like an old-fashioned blackboard one) where we discuss the basic details. The week thereafter we will discuss how to implement a neural network for solving the same problem as discuss with Boltzmann machines. The week of May 1-5 will also be our last regular lecture week. The weeks thereafter will be dedicated to a summary of what has been done and just project work.
Best wishes to you all,
Morten and Øyvind
Dear all, since most of you seem to have chosen the project on deep learning, the remaining lectures will focus on generative models like Boltzmann machines as well as the implementation of neural networks.
The lecture slides as a jupyter-notebook for the week of April 17-21 are at https://github.com/CompPhysics/ComputationalPhysics2/blob/gh-pages/doc/pub/week13/ipynb/week13.ipynb
This week we will discuss the basics of Boltzmann machines.
For those of you interested in the variant with fermions only, the lecture notes at http://compphysics.github.io/ComputationalPhysics2/doc/LectureNotes/_build/html/vmcdmc.html#vmc-for-fermions-efficient-calculation-of-slater-determinants
Best wishes to you all
Morten and Øyvind.
Dear all, we hope you have had an enjoyable break.
The aim this week is to present and discuss five possible variants for project 2, see https://github.com/CompPhysics/ComputationalPhysics2/tree/gh-pages/doc/Projects/2023/Project2 for the various project files.
There are five topics at present
1) Time-dependent Hartree-Fock theory
2) VMC with fermions, continuation of project 1
3) VMC with deep learning, also continuation of project 1 but now the trial wave function is replaced with a neural network
4) Quantum computing using the variational quantum eigensolver approach
5) For those more familiar with many-body methods like coupled-cluster theory, there is a project variant which deals with the development of a coupled-cluster theory code.
All these alternatives will be presented and discussed during the lecture on April 13. We continue thereafter with a discussion of th...
Dear all, I am sorry for changing time again, but Øyvind needs to change the lab time back to 1615-19 instead of 1415-17 as announced earlier. Furthermore, the person we were supposed to interview for a position changed the travel plans for tomorow.
This means that lectures and lab tomorrow will be at our regular time. I am sorry for this last moment changes.
We have thus
1) Lecture at 1415-16 and
2) lab 1615-19
We will wrap our discussion on project 1 (we are also flexible with the deadline, as always) and discuss parallelization and vectorization strategies. Many of you are already familiar with these topics and Monte-Carlo approaches like the variational MC approach we use in the project are normally extremely trivial to parallelize.
During the lecture tomorrow we will review basic elements needed in the project and discuss how we parallelize the codes. Most of the examples are in C++. The slides are at for exam...
Dear all, thx to those who could attend at today's changed lecture time 1015-12.
For those of you who could not attend, the link to the video of the lecture is at https://youtu.be/zbxvPJ5YxR8
and the whiteboard notes are at https://github.com/CompPhysics/ComputationalPhysics2/blob/gh-pages/doc/HandWrittenNotes/2023/NotesMarch23.pdf
The lab begins at 1415 today.
Best wishes to you all and remember that next week we have also to anticipate the lecture, this time to 1215-14, with lab thereafter.
Hope that is ok for you all and best wishes
Morten and Øyvind
Dear all, we hope you are all doing well. here follows an update with plans for this week and next week (week before the easter break).
Note also that we need to change the lecture time this week and next week due to Morten chairing a position search. Unfortunately this collides with our usual lecture times this week and next week. The candidates are giving their seminars and interviews during our regular lecture time.
The changes are as follows, and I hope you can join the zoom sessions, or just watch the lecture video
1) Thursday March 23: lecture at 1015am-12pm and lab from 14.15-1700 (alternatively longer, depends on Øyvind). Tomorrow we discuss blocking and resampling with code examples, see jupyter-notebook at https://github.com/CompPhysics/ComputationalPhysics2/blob/gh-pages/doc/pub/week9/ipynb/week9.ipynb
2) Thursday March 31
Lecture at 1215-14 and lab 14-17 (or longer)....
Dear all, we hope you've had a great start of the week. Tomorrow, March 2, we plan to have lab only from 215pm to 7pm, at our usual location FØ397. The aim is to try to wrap up standard VMC with importance sampling for systems of non-interacting bosons in 1, 2 and 3 dimensions, and hopefully get started with the optimization part. The week thereafter,March 6-10, we will replace most of the lab with lectures only, with an emphasis on gradient methods tailored to VMC calculations.
Best wishes to you all,
Morten and Øyvind
Dear all, we hope you have had a great start this week. Here follows a short update about the plans for the lectures tomorrow. We will start discussing about the optimization of the variational parameters that define the trial wave function. This will us into the world/jungle of gradient descent method, methods whom many of you are already familiar with from courses like FYS-STK4155. Here we will adapt these methods to the optimization of the ground state energy in order to be able to launch large scale Monte Carlo calculations. The slides for this week are at http://compphysics.github.io/ComputationalPhysics2/doc/pub/week5/html/week5-reveal.html
Several of you have also asked for a possible change of lab time, that is from Thursdays 4pm-7pm.
Depending on available space we could have the labs on
1) Tuesdays 9-12pm or 2pm-5pm
2) Wednesdays 9-12pm or 2pm-5pm.
Feel free to send me an email at mhjensen@uio.uio with your preferred choice. De...
This week we will finalize our discussion on importance sampling with several coding elements. We will start with optimization studies as well.
Dear all, we hope you are doing well. This week we will have a regular zoom lecture, which will also be recorded. After having discussed and derived the Metropolis algorithm last week, we will now focus on implementing the Metropolis-Hastings variant of the algorithm by introducing what is called Importance sampling.
We will discuss coding during the lectures as well as the mathematics behind importance sampling. The plan is to discuss
- Markov Chain Monte Carlo
- Metropolis-Hastings sampling and Importance Sampling
- Fokker-Planck and Langevin equations
The lecture material with jupyter-notebook and more is at http://compphysics.github.io/ComputationalPhysics2/doc/web/course, just scroll down to the relevant week.
After the lecture, we will have our regular lab session.
Best wishes to you all and don't hesitate to ask questions.
Morten and Øyvind
Dear all,
below you will a questionnaire about finding possible partners for group work in FYS4411/9411,
see
https://docs.google.com/forms/d/1lVESglIY4Jx_CXs4jeTnVX3d7kP5j72F6lABnzt90j0/edit
Else, you should all have access to the lecture video from Thursday the 26th. Next week we will discuss in detail how to build a variational Monte Carlo code and we will discuss the theory of Markov chains and derive the Metropolis algorithm. The teaching material is at http://compphysics.github.io/ComputationalPhysics2/doc/pub/week2/html/week2-reveal.html
Project one is also available from https://github.com/CompPhysics/ComputationalPhysics2/tree/gh-pages/doc/Projects/2023/Project1
Best wishes to you all,
Morten and Øyvind
Dear all,
sorry today for the confusion about in person lectures. All lectures, except for March 9, will be via zoom.
For those of you who could not attend, here's the link to the video of today's lecture, see https://youtu.be/2wg5rDJ7bek
The handwritten notes are at https://github.com/CompPhysics/ComputationalPhysics2/blob/gh-pages/doc/HandWrittenNotes/2023/NotesJan262023.pdf
Else, all teaching material is available via the GitHub site at https://github.com/CompPhysics/ComputationalPhysics2
Feel free to download this material and use it as your own.
I will also send you all a questionnaire about your background and interests in teaming up with other people for work on the projects.
Next week we will continue our discussion on how to build a VMC code, discuss the Metropolis algorithm and more. The slides for next week are at http://compphysics.github.io/ComputationalPhysics2/doc/web/course, just scroll down to the slides ...
Dear all, welcome to a new semester and FYS4411/9411.
The course has its emphasis on Monte Carlo with and without deep learning methods for finding optimal trial wave functions.
All lectures will be recorded and we use zoom with the following zoom link for lectures and the lab sessions
Zoom link for fys4411/9411
https://msu.zoom.us/j/6424997467?pwd=d0xQUDZmaU00T1Job1J3RnVuL3l6UT09
Meeting ID: 642 499 7467
Passcode: FYS4411
Furthermore, all educational material is at the GitHub link https://github.com/CompPhysics/ComputationalPhysics2
Welcome to a new semester and we hope you will enjoy FYS4411/9411.
Morten and Øyvind