Develop a new module on Monte Carlo methods

[labarba]

In email conversation, @bknaepen said his curriculum included Monte Carlo, a topic we've not covered in numerical-mooc. Recently, I talked with @afeiguin (Adrian) who teaches a computational physics course at Northeastern University:
http://www.northeastern.edu/afeiguin/phys5870/

I introduced Adrian to Jupyter and he is enthusiastic about adopting Python for his course. We have an opportunity here to collaborate in the development of a new module on Monte Carlo methods — @IanHawke was also interested, and he's got some Python code for one of the NGCM modules.

Let's start a discussion here about how a module on MC might look like and how to pull our resources together to make it happen!

[IanHawke]

There's some simple Python MC code at https://github.com/ngcm/training-public/tree/master/FEEG6016%20Simulation%20and%20Modelling but the details of some of the exercises with LJ potentials aren't there. Also, I haven't put the solutions up (can email them).

[bknaepen]

Tomorrow I will see if our local Monte Carlo guru is willing to jump on board and provide some help...

[bknaepen]

Dear all,

Bortolo Mognetti and I have made some progress on some of the content. The status is:

1. We have written a working code to discuss the Ising model in 2D. It will give us the opportunity to introduce the Metropolis algorithm. This part now needs to be written in notebook format but we have the material ready in .py files and on ‘traditional’ paper.

2. We have also significantly discussed the content of one (or two) notebooks about Brownian motion. We want to use this as an introduction to random numbers and to the generation of arbitrary distributions of random numbers. We also want to include in that part the resolution of a financial stochastic model like the Black-Scholes model. This part needs further programming but the outline is there.

Teaching of this MC material starts around 1st of February so will likely have a draft ready by the end of 2015 (hopefully the Ising model stuff sooner)…

[afeiguin]

I will be covering the Ising model next week, we can probably exchange
notes. I am still trying to decide whether I will cover critical slowing
down and cluster updates (Wang-Landau).

On 11/16/15 9:42 AM, bknaepen wrote:

Dear all,

Bortolo Mognetti and I have made some progress on some of the content.
The status is:

1. We have written a working code to discuss the Ising model in 2D. It
   will give us the opportunity to introduce the Metropolis algorithm.
   This part now needs to be written in notebook format but we have the
   material ready in .py files and on ‘traditional’ paper.

2. We have also significantly discussed the content of one (or two)
   notebooks about Brownian motion. We want to use this as an
   introduction to random numbers and to the generation of arbitrary
   distributions of random numbers. We also want to include in that part
   the resolution of a financial stochastic model like the Black-Scholes
   model. This part needs further programming but the outline is there.

Teaching of this MC material starts around 1st of February so will
likely have a draft ready by the end of 2015 (hopefully the Ising
model stuff sooner)…

—
Reply to this email directly or view it on GitHub
#137 (comment).

[bknaepen]

It looks like you are way ahead of us :-). I will take a look later and forward the information to Bortolo...

[IanHawke]

What audience are we aiming at with these lessons? The first few chapters used engineering-style examples. The MC notebooks I have use computational chemistry examples (thanks to the colleagues that provided the background). I tend to think of Ising as more physics-oriented.

Should we go with a single, motivating example (as in the phugoid case), or a range of examples?

[afeiguin]

I am aiming at a physics audience. I will do the classical gas in 1d,
Ising in 2D, and cluster updates. After that I will include quantum
fluctuations and do exact diagonalization of the fully quantum
Heisenberg model, and that will be the end of this course. I have even
prepared lectures on world-line quantum MC for a more advanced course
(not this time), and also prepared lectures on comp. chem.-like
problems, using the variational principle to solve the H atom and H2
molecule using gaussians, do you have anything like that in your course?

On 11/16/15 5:32 PM, Ian Hawke wrote:

What audience are we aiming at with these lessons? The first few
chapters used engineering-style examples. The MC notebooks I have use
computational chemistry examples (thanks to the colleagues that
provided the background). I tend to think of Ising as more
physics-oriented.

Should we go with a single, motivating example (as in the phugoid
case), or a range of examples?

—
Reply to this email directly or view it on GitHub
#137 (comment).

[IanHawke]

I've got MD simulations of H2O and Metropolis-Hastings of Lennard-Jones,
but not nearly enough time to cover much else (although the Hartree-Fock
section needs updating, and better examples put in there).

I would say that a physically-minded audience is the direction to go in for
this section, agreed.

On Tue, Nov 17, 2015 at 2:00 PM, afeiguin notifications@github.com wrote:

I am aiming at a physics audience. I will do the classical gas in 1d,
Ising in 2D, and cluster updates. After that I will include quantum
fluctuations and do exact diagonalization of the fully quantum
Heisenberg model, and that will be the end of this course. I have even
prepared lectures on world-line quantum MC for a more advanced course
(not this time), and also prepared lectures on comp. chem.-like
problems, using the variational principle to solve the H atom and H2
molecule using gaussians, do you have anything like that in your course?

On 11/16/15 5:32 PM, Ian Hawke wrote:

What audience are we aiming at with these lessons? The first few
chapters used engineering-style examples. The MC notebooks I have use
computational chemistry examples (thanks to the colleagues that
provided the background). I tend to think of Ising as more
physics-oriented.

Should we go with a single, motivating example (as in the phugoid
case), or a range of examples?

—
Reply to this email directly or view it on GitHub
<
#137 (comment)
.

—
Reply to this email directly or view it on GitHub
#137 (comment)
.