|
David Kipping
@
david_kipping
Manhattan, NY
|
|
Exoplanets & Exomoons, Cool Worlds Lab, Columbia University
|
|
|
7.054
Tweetovi
|
119
Pratim
|
5.399
Osobe koje vas prate
|
| Tweetovi |
|
David Kipping
@david_kipping
|
4. velj |
|
Yes!
|
||
|
|
||
|
David Kipping
@david_kipping
|
4. velj |
|
I’d say that’s actually too general because it’s two-way. I want the one way transition into a lower energy state only....
|
||
|
|
||
|
David Kipping
@david_kipping
|
4. velj |
|
Yes along those lines but something formally defined in a bibliographic source ideally!
|
||
|
|
||
|
David Kipping
@david_kipping
|
4. velj |
|
Perhaps that example is not the best but I think there is a more general definition to be had regarding phase transitions
|
||
|
|
||
|
David Kipping
@david_kipping
|
4. velj |
|
I’m struggling to find a physics-based formal definition of condensation. Most online describe liquid-gas phase transition. But I want a more general one, for example how the fundamental forces condensed out of a unifying force in the early Universe. Is there something like that?
|
||
|
|
||
|
David Kipping
@david_kipping
|
3. velj |
|
Most humans will lose consciousness at 5g twitter.com/OGMowglii/stat…
|
||
|
|
||
| David Kipping proslijedio/la je tweet | ||
|
James Davenport, PhD
@jradavenport
|
3. velj |
|
Video is now live!! Check it out! twitter.com/jradavenport/s…
|
||
|
|
||
|
David Kipping
@david_kipping
|
3. velj |
|
I was in the live chat!
|
||
|
|
||
| David Kipping proslijedio/la je tweet | ||
|
guy fieri debord
@bispectral
|
2. velj |
|
I might be the last astronomer to see this but I am in TEARS at the MIRI logo pic.twitter.com/0UZVTNZ8NS
|
||
|
|
||
|
David Kipping
@david_kipping
|
3. velj |
|
Deep cut
|
||
|
|
||
|
David Kipping
@david_kipping
|
3. velj |
|
Had a blast with Jim, make sure you check out this episode and his wonderful channel! twitter.com/jradavenport/s…
|
||
|
|
||
|
David Kipping
@david_kipping
|
3. velj |
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
I agree but I think you’re describing a summary statistic there rather than the sampler itself. I think the conversation keeps jumping between the two which is making it low SNR. This would just be better to discuss in person next time.
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
I think it depends on the implementation. For the popular MultiNest code it’s just an issue of turning on the flag nest_pWrap=1. I would think periodic parameters would be easy to deal with irrespective of the sampling algorithm just by using numerical tricks like truncation.
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
Awesome to be included in this list of educational @YouTube channels! Especially to be put in the list next to outrageously good @Kurz_Gesagt! collegeinfogeek.com/educational-yo… pic.twitter.com/WS8ps6IYZf
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
Yes i usually see something different. I can’t make a plot though - away from my computer. But! If you don’t have any such bias with whatever sampler you’re using then I don’t think there’s any good reason to reparameterize to the Lagrangian-esque components anyway
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
No we're talking about different effects. LW71 describe how using a summary statistic of a mean/median is positively biased when applied to a posterior near a BC. This is a different effect where the posterior is additionally skewed. Easier just chat next time I see you about it!
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
This would go away with an infinitesimal proposal size, and maybe could be solved with a better sampler, I think the reparameterization in the OP indeed solves it too. But you don’t really have this issue to fix with nested sampling which I appreciated when I switched!
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
The problem I found nested sampling avoids is that if my mcmc makes a proposal from e=0.01 with a Gaussian proposal of 0.02, there’s a decent chance it will try a negative e that is forbidden. So it’s biased toward accepting positive jumps here, which skews the posterior.
|
||
|
|
||
|
David Kipping
@david_kipping
|
2. velj |
|
I actually don’t use that reparameterization since I stopped using Markov chains. Nested sampling doesn’t suffer from artificially positively biased posteriors against boundary conditions as troubles MCMC.
|
||
|
|
||