/sci/ - Science & Math » Thread #13111391

176KiB, 800x715, 1512596450327.jpg

View Same Google iqdb SauceNAO

Anonymous Mon 10 May 22:41:24 2021 No.13111391 View Reply Original Report

Quoted By: >>13111420 >>13113421

>essay on strengths and weaknesses of Markov Chain Monte Carlo models due in 24hr
>Reading about them for the first time
I'm never gonna use this in industry, also help pls

Anonymous

Anonymous Mon 10 May 2021 22:48:16 No.13111420 Report

Quoted By:

>>13111391
you are going to have to retake this course next year anon

Anonymous

Anonymous Tue 11 May 2021 00:17:27 No.13111846 Report

Quoted By: >>13113348

MCMC is not a model, but a collection of inference algorithms used to sample from posterior distributions.
Weakness: MUCH SLOWER than variational inference, in metropolis hastings you have to define the step size, in gibbs sampling you need to be able to sample from each partial conditional distribution. Difficulty in determining convergence to the target distribution.
Strengths: More accurate than variational inference. Unbiased estimation.

I use them all the time in my projects.

Anonymous

Anonymous Tue 11 May 2021 07:47:07 No.13113348 Report

Quoted By:

>>13111846
thanks lad

Anonymous

Anonymous Tue 11 May 2021 08:09:21 No.13113413 Report

Quoted By:

I have a course on them right now actually. We’re currently implementing a hybrid MCMC algorithm in a project. Seems highly applicable

Anonymous

Anonymous Tue 11 May 2021 08:11:53 No.13113421 Report

Quoted By:

>>13111391
>Weakness
A Markov Chain is only as strong as its weakest link

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches