These are chat archives for libmir/public

15th
May 2016
Joseph Rushton Wakeling
@WebDrake
May 15 2016 14:25
Re the proposal to add references for @wilzbach's project to the wiki at https://github.com/libmir/mir/wiki ... I cannot find said wiki ;-)
The URL just forwards on to the usual project GitHub page
Ilya
@9il
May 15 2016 14:26
I turned off the wiki because this is not filled
Sebastian Wilzbach
@wilzbach
May 15 2016 14:26
@9il do we have a backup of the wiki?
Ilya
@9il
May 15 2016 14:27
Yes, see right now, it is availble
Sebastian Wilzbach
@wilzbach
May 15 2016 14:27
great, should I remove all NumPy pages temporarily?
Ilya
@9il
May 15 2016 14:28
Yes, but i don’t know how to save them
Sebastian Wilzbach
@wilzbach
May 15 2016 14:28
No worries - it's git ;-)
okay - everything is removed. I made a local backup, it's in the git history and in a (hidden) "numpy" branch ;-)
Ilya
@9il
May 15 2016 14:31
ok, thanks
Sebastian Wilzbach
@wilzbach
May 15 2016 14:31
probably you should update this page then?
Ilya
@9il
May 15 2016 14:31
What page?
sorry link didn't go through
Joseph Rushton Wakeling
@WebDrake
May 15 2016 14:33
@9il so is the wiki back to stay? I can submit proposed changes there ... ?
Or @wilzbach would you prefer that I just send you reference lists for you to add? (I'm fine either way)
Sebastian Wilzbach
@wilzbach
May 15 2016 14:33
@WebDrake whatever is more convenient for you ;-)
Ilya
@9il
May 15 2016 14:34
@WebDrake Do you need edit access for the wiki?
Sebastian Wilzbach
@wilzbach
May 15 2016 14:34
@9il everyone can edit the wiki
Ilya
@9il
May 15 2016 14:35
oh, ok
Joseph Rushton Wakeling
@WebDrake
May 15 2016 14:35
Ah, cool, wasn't sure if I needed committer privileges
In that case I'll try to add some references later today
Thanks!
BTW the Karney (2013) paper already referened there looks interesting, I wasn't aware of that
Sebastian Wilzbach
@wilzbach
May 15 2016 14:39
@WebDrake thanks - list is pretty out-of-date and I will also update this in next days ;-)
Joseph Rushton Wakeling
@WebDrake
May 15 2016 18:27
Favour to ask -- would be really interested in people's feedback on https://github.com/WebDrake/dxorshift in general and WebDrake/dxorshift#1 in particular
Ilya
@9il
May 15 2016 18:34
Does it have Phobos API?
Joseph Rushton Wakeling
@WebDrake
May 15 2016 18:37
They all satisfy isUniformRNG constraints, but they are all input ranges with postblit disabled
I thought I'd take advantage of implementing these RNGs to try out that approach and see how people find them in practice
The open PR is for a wrapper to facilitate using them in @safe code
Joseph Rushton Wakeling
@WebDrake
May 15 2016 22:39
@wilzbach I've added some references to the wiki
Sebastian Wilzbach
@wilzbach
May 15 2016 22:40
Saw it. Thanks a lot :)
Joseph Rushton Wakeling
@WebDrake
May 15 2016 22:40
Particularly interesting for you to look at may be Thomas et al. (2007) -- compares lots of different ways of generating normal variates
Their methodology may be interesting in terms of testing quality of output
Sebastian Wilzbach
@wilzbach
May 15 2016 22:42
Will have a look tomorrow / next days. Thanks again. I really hope I can make some change with my project
Joseph Rushton Wakeling
@WebDrake
May 15 2016 22:43
Ping me (here or via email) if there's any direct advice or assistance you need :-)
BTW one particular note about Ziggurat algorithm
AIUI it can be used for any monotone decreasing distribution, and hence by extension any symmetric unimodal distribution
Which means it should be possible to use it not only for normal and binomial distributions, but also e.g. exponential and pareto
Joseph Rushton Wakeling
@WebDrake
May 15 2016 22:48
I've never come across it being used for the last of these, so it would be interesting to try
Pareto is fairly straightforward though to do just on the basis of transforming U(0, 1) via the CDF
But I'm not sure what quality of statistics results from that
Joseph Rushton Wakeling
@WebDrake
May 15 2016 22:55