Edge Rewrite
// HTMLRewriter · presentation

This page was redesigned at the edge.

Cloudflare fetched the original article and streamed it through HTMLRewriter to apply an entirely new visual system without rebuilding the source page.

// request.cf · coarse context

A page that knows where it met you.

Only coarse request metadata is shown. This demo does not display or persist visitor IP addresses.

Country
US
Cloudflare location
CMH
Connection
HTTP/2
Language
Not provided

Ray ID: a21fdafe5aba2a83

Jump to content

Talk:Random search

Page contents not supported in other languages.
Add topic
From Wikipedia, the free encyclopedia
Latest comment: 10 years ago by 123.16.157.60

There is a paper by Wallace about using the Walsh-Hadamard transform to generate random numbers with a Gaussian distribution. The Gaussian distribution is directly related to picking a point uniformly at random on the surface of a hypersphere. You can use the Walsh-Hadamard transform together with random permutations to generate random points on the hypersphere. Both the WHT and random permutations leave vector length unchanged. A sequence of random permutations and WHT's does interesting things. Or earlier even: https://archive.org/details/bitsavers_mitreESDTe69266ANewMethodofGeneratingGaussianRando_2706065  Preceding unsigned comment added by 123.16.157.60 (talk) 00:51, 26 March 2016 (UTC)Reply


In this article, it states that random search works by searching in the hypersphere surrounding the current position in the solution space. While this may true in the method of Rastrigin, this article is about the family of search algorithms under the umbrella of random search. This method of searching nearby solutions ought to be called something like random local search, so as to permit uniform global random search to exist under the umbrella of random search. Uniform global random search meaning that every possible solution in the search space has equal likelihood of being selected in a given iteration. -- Sohrab


I agree with Sohrab. Even though Rastrigin used the term "Random Search", the algorithm described here would nowadays be called a (black-box) Hill-Climber. And nowadays, Random Search just means drawing independent samples from the space, uninformed by previous samples. Some arbitrary examples:

So, I would propose to move some of the contents of this page to the wikipedia Hill_Climbing page, and on this page, instead describe Random Search in the modern sense. - Jamesmichaelmcdermott