Questions tagged [quasi-monte-carlo]

quasi monte carlo is a technique for doing monte carlo integration and other monte carlo simulations, replacing the usual pseudo random sequence with a low-discrepancy sequence. This can be seen as a general trick to lower the variance introducing dependency into the random number sequence.

31 questions
Filter by
Sorted by
Tagged with
1 vote
19 views

Does it make sense to compare $KL(q_1|p)$ vs $KL(q_2|p)$ where $q_1$ and $q_2$ are empirical distributions?

I simulated x samples of N(0,1) using Sobol sequence vs 100x samples of N(0,1) peuso-random number. I chopped up simulations to 100 samples for peudo-random number. I found about 50 times the KL ...
• 656
10 views

Quasi Monte Carlo: Is not grid search the ultimate low discrepancy sequence?

I realize that subsequences won't be low discrepancy, but if I know how many samples I want and that I am operating on the unit hypercube, why not just grid search as the low discrepancy sequence in ...
• 21
1 vote
25 views

Generation of completely uniformly distributed sequences

In quasi-Monte Carlo, there is a rather strong notion "completely uniformly distributed" sequences, which somehow mimics independence and is described, for example, at the end of this post. ...
• 293
118 views

Is there a Quasi-Monte Carlo variant of the Metropolis-Hastings algorithm?

If we run the Metropolis-Hastings algorithm for a target distribution $\mu$ with proposals from a quasi-Monte Carlo sequence $(y_n)_{n\in\mathbb N}$ (such as a Sobol sequence) and the generated chain ...
• 293
1 vote
47 views

"Variance" estimate for quasi-Monte Carlo

Problem Setting I am playing with a toy example and I would like to better understand the variance results that I get when using low-discrepancy sequences versus random values. I have independent and ...
• 111
616 views

How to calculate Quasi-Monte Carlo integration error when sampling with Sobol's sequence?

My understanding is that QMC integration using random sampling will converge with $O(\frac{1}{\sqrt{n}})$, while using Sobol's sampling will converge with $O(\frac{(\log{n})^d}{n})$. However I'm ...
• 173
1 vote
104 views

What are some methods to choose a $n$ for Quasi Monte Carlo Integrations?

When studying "simple" Monte Carlo integration methods, such as Hit or Miss, Crude , Importance Sampling, etc. A common problem for first time learners is to choose a number $n$ of points ...
146 views

Quasi Monte Carlo estimation of logit-normal density integrals

I am considering the integral $$I(y \mid \mu, \sigma) = \int_y^1 \frac{\exp \left\{ \frac{-1}{2\sigma^2}[\textrm{logit}(x)-\mu)]^2 \right\}}{\sigma \sqrt{2\pi} (1-x)}\textrm{d}x$$ which for $y=0$ is ...
• 2,907
146 views

Quasi-random sequence with discrete variables with differing number of levels

My question is probably worded incorrectly but here it is: I have (say) 3 discrete random variables: x1: has 15 levels (uniform pdf for simplicity) x2: has 3 levels (uniforms) x3: has 4 levels (...
• 21
704 views

How to get normally distributed Quasi-random numbers

I've been trying to use the chaospy package to get quasi-random numbers for a Monte Carlo simulation. The dimensions need to be 365×5000 (but can be up to 2190×5000). When I pull a sample using ...
• 131
1 vote
183 views

Stratified sampling / QMC simulation for compound Poisson rv

I have a rv $X$ of the form $$X=\sum_{i=1}^N Y_i,$$ where $N$ is a discrete rv (often, but not always, Poisson) and $Y_1,\ldots,Y_N$ are continuous random variables, iid and independent from $N$. I ...
• 237
85 views

Inverse transform method on MCMC generated uniform draws

I understand that it sounds like why would anyone do this, but are there any references that use the inverse transform method to draw correlated samples from a distribution $F$ using MCMC samples from ...
• 15.6k
1 vote
24 views

Are subsequences of low discrepancy sequence also low discrepancy? [closed]

Given a low discrepancy sequence x1 ... xN, lets say I randomly select a subsequence x100 ... x200, will this subsequence have low discrepancy? Will those points "fill" up the space uniformly? If not ...
• 11
73 views

How to apply linear transformation? [closed]

In this post, Martin Roberts mentioned that: ... to convert to a range of [-1,1], simply apply the linear transformation x:=2x+1. The result is (-0.361655, -0.657913, -0.900599) (-0.72331, 0.684174, ...
1 vote
190 views

Can/should one generate Ginibre ensembles of random matrices using low-discrepancy normal variates--and if so. how?

I've been generating (via Mathematica) series of $4 \times 4$ "random density matrices with respect to Bures (minimal monotone) measure" https://arxiv.org/abs/0909.5094 [eq. (24)] and testing certain ...
1 vote
1k views

Sobol Sensitivity Analysis

I want to use Sobol SA with Sobol sampling to find the most influential parameters on the energy consumption of a pilot building. I have 40 input variables (building characteristics) that some have ...
53 views

• 423
2k views

How to estimate the accuracy of an integral?

An extremely common situation in computer graphics is that the colour of some pixel is equal to the integral of some real-valued function. Often the function is too complicated to solve analytically, ...
• 2,750
7k views

Best method for transforming low discrepancy sequence into normal distribution?

I've been using low discrepancy sequences for a while for Uniform Distributions, as I've found their properties useful (mainly in computer graphics for their random appearance and their ability to ...
• 253
5k views

Halton sequence vs Sobol' sequence?

From an answer in a previous question, I was directed toward the Halton sequence, for creating a set of vectors that covered a uniform sample space fairly evenly. But the wikipedia page mentions that ...
• 5,453
2k views

Increase the sample size of a Latin Hypercube study?

I want to create a climate model ensemble, testing 5 parameters (real, uniformly distributed between two values), using a latin hypercube approach. The problem is that I'm not sure how many ...
• 5,453