Re: a few questions about random numbers

Thread: Re: a few questions about random numbers

Search this list only Search all lists
Re: a few questions about random numbers
United States	2008-03-04 09:34:59

Dear Seth, Thank you very much. I tried to set the random-seed in the setup procedure but the model still runs differently every time. Is there a particular number of random-seed that I need to designate?? Appreciated much, Nicole On Mon, Mar 3, 2008 at 12:27 PM, Seth Tisue < sethtisue.net">sethtisue.net> wrote: >>>>> "Nicole" == Nicole Li < nicole.ykli%40gmail.com" target="_blank">nicole.ykligmail.com> writes: Nicole> (A) Besides the Netlogo primitives with "random" in their names Nicole> (e.g., random, random-float, random-normal), what other Netlogo Nicole> operations/ commands will make random choices? The manual said Nicole> that the primitives: one-of, n-of, sprout, and downhill are Nicole> all governed by random seed. Are there any others??? We don't have a list, but the individual entries in the NetLogo Dictionary should say if a primitive does something random. Remember also that any operation involving an agentset may use random numbers, since agentsets are always in random order. So that means most models make pervasive use of randomness throughout. Nicole> (B) I'm a bit unclear about the primitive of Nicole> "with-local-randomness". Besides the examples in the manual, Nicole> are there any models in Netlogo library using Nicole> "with-local-randomness" and so I can learn more how to use Nicole> it??? There aren't. It's not a primitive a model would normally use. If you're not sure if you need it, you almost certainly don't. Nicole> (C) Let's say I want to create reproducible model runs. My Nicole> model has a number of agentsets. Each agentset use different Nicole> "random" primitives, e.g., the 1st agentset has Nicole> "random-normal", the 2nd agentset has "sprout" and the 3rd Nicole> agentset has "one-of". If I want to create reproducible Nicole> model runs, is that: (1) I need to record the random seed for Nicole> the first run of the model that I'd like to reproduce, (2) I Nicole> need to use "with-local-randomess"; for each agentset (???), and Nicole> (3) I need to use the same random seed as in (1) for the second Nicole> run? As such, the model will use the same random seed for all Nicole> the "random" primitives across the agentsets. And the model's Nicole> outputs will be the same in the first run and the second Nicole> run. Is that correct? If you want to a reproducible model run in NetLogo, all you need to do is set the random seed beforehand. You don't need to do anything else special. Try it! Hope this helps, -- Seth Tisue / http://tisue.net lead developer, NetLogo: http://ccl.northwestern.edu/netlogo/ -- Nicole yee-key LI Doctoral Student Department of Communication Science and Disorders University of Pittsburgh __._,_.___ Messages in this topic (3) Reply (via web post) \| Start a new topic . __,_._,___
Re: a few questions about random numbers
United States	2008-03-04 11:24:54

>>>>> "Nicole" == Nicole Li < nicole.ykli%40gmail.com">nicole.ykligmail.com> writes: Nicole> Dear Seth, Thank you very much. I tried to set the random-seed Nicole> in the setup procedure but the model still runs differently Nicole> every time. Is there a particular number of random-seed that I Nicole> need to designate?? Your model is probably doing something that depends on external factors. For example, are you using EVERY? (see my reply to Gary An) I can't rule out that you've found a bug in NetLogo, but before suspecting that, you should exclude the possibility that there's something special in your code that cause the results to vary for other reasons than varying the random seed. -- Seth Tisue / http://tisue.net lead developer, NetLogo: http://ccl.northwestern.edu/netlogo/ __._,_.___ Messages in this topic (4) Reply (via web post) \| Start a new topic . __,_._,___

[1-2]

about | contact Other archives ( Real Estate discussion Medical topics )