4 ms·
Something I find is underemphasised in this type of material is the equivalence between computing the value of an integral and the sum of functions of random va
by kqr 6y ago
Something I find is underemphasised in this type of material is the equivalence between computing the value of an integral and the sum of functions of random variables. This is a very useful intuition for framing various problems in efficient ways.
This article brings it up, but only as a "look at what you can do with MC methods" and not "here's this fundamental relation that's really important."
There's lots of useful understanding about importance sampling and related techniques that seem obvious when you look at it as an integration problem.
- Bootvis 6y agoCould you point to a resource that explains method well?
- kqr 6y agoUnfortunately my only exposure to this is a University course I took called Computer Intensive Methods in Mathematical Statistics[1]. It was a really good course, but at the time I was not in a mental space where I could get the most out of it, so I'm kind of hoping someone else can point to a solid textbook or something. [1]: https://www.kth.se/student/kurser/kurs/SF2955?l=en https://www.kth.se/student/kurser/kurs/SF2955?l=en
- nbernard 6y agoI liked the book "Statistical Mechanics: Algorithms and Computations" by Werner Krauth. It's oriented toward physics, but IIRC he links Monte Carlo simulations to the path integral formalism. There is also a MOOC (same title, same author) on Coursera that I found interesting. It doesn't not go to the same depth as the book however. They are more complementing each other, the book for theory, the MOOC for implementation exercises.
- sitkack 6y agothis looks wonderful https://www.coursera.org/learn/statistical-mechanics https://www.coursera.org/learn/statistical-mechanics
- oyoman 6y agoI don't know if I can plug-in here some of the materials I prepared some years ago. It might help to bridge/connect the use of Monte Carlo techniques and its relations with integration: https://github.com/ChristopheRappold/MCandPython/blob/master/session2.ipynb https://github.com/ChristopheRappold/MCandPython/blob/master... Also just to mention something else that is very interesting related with MC : "quasi-Monte Carlo techniques" (small example in https://github.com/ChristopheRappold/MCandPython/blob/master/session3.ipynb https://github.com/ChristopheRappold/MCandPython/blob/master... ). For people interested in MC, just take a look to those keywords.