UmbralRaptor changed the topic of #kspacademia to: https://gist.github.com/pdn4kd/164b9b85435d87afbec0c3a7e69d3e6d | Dogs are cats. Spiders are cat interferometers. | Космизм сегодня! | Document well, for tomorrow you may get mauled by a ネコバス. | <UmbralRaptor> … one of the other grad students just compared me to nomal O_o | <ferram4> I shall beat my problems to death with an engineer.
<bofh>
egg|zzz|egg: HUH so staring at that eccentricity perturbation plot of Bop I'm immediately reminded of how it looks superficially familiar but more complicated than a plot of the equation of time.
<bofh>
also yuck, integrating the relevant variational eq'n is an unholy pain in the arse
_g__k has joined #kspacademia
_g__k has left #kspacademia [#kspacademia]
e_14159 has quit [Ping timeout: 195 seconds]
e_14159 has joined #kspacademia
Technicalfool is now known as TechnicallySleeping
TechnicallySleeping is now known as Technicalfool
Technicalfool is now known as TechnicallySleeping
Majiir is now known as Snoozee
icefire has quit [Read error: Connection reset by peer]
* Qboid
gives bofh, Fiora, et al. an expressed function generator
egg|zzz|egg is now known as egg|afk|egg
egg|phone|egg has quit [Read error: Connection reset by peer]
egg|phone|egg has joined #kspacademia
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Read error: -0x1: UNKNOWN ERROR CODE (0001)]
egg|phone|egg has joined #kspacademia
egg|cell|egg has quit [Read error: Connection reset by peer]
egg|phone|egg has quit [Read error: Connection reset by peer]
egg|phone|egg has joined #kspacademia
<whitequark>
!wpn egg
* Qboid
gives egg a ship
<egg|work|egg>
!wpn whitequark
* Qboid
gives whitequark a war-surplus diffeomorphism
<egg|work|egg>
whitequark: how are the cats doing
<whitequark>
q. fine
<whitequark>
the kitten is now healthy and also far less annoying
<whitequark>
the cat has learned to hiss at and beat the kitten when the kitten is pestering the cat too much so the cat now has far less of a neurosis issue
<egg|work|egg>
yay
<egg|work|egg>
whitequark: have you found a gas mask for cats?
<whitequark>
nope
<whitequark>
havent been particularly actively looking
<egg|work|egg>
whitequark: also, icymi Qboid now counts down to rocket launches
<egg|work|egg>
!wpn котя
* Qboid
gives котя a pi teledildonics
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Read error: Connection reset by peer]
egg|phone|egg has joined #kspacademia
egg|cell|egg has quit [Read error: Connection reset by peer]
egg|cell|egg has joined #kspacademia
egg|phone|egg has quit [Read error: Connection reset by peer]
egg|cell|egg has quit [Read error: Connection reset by peer]
<Qboid>
egg|work|egg: Description: A SpaceX Falcon 9 rocket will launch the 14th Dragon spacecraft on the 12th operational cargo delivery mission to the International Space Station.
<egg|work|egg>
Thomas: kd next is buggy
<egg|work|egg>
Thomas: also Qboid isn't counting down properly as the launch approaches
<Qboid>
Thomas: Description: A SpaceX Falcon 9 rocket will launch the 14th Dragon spacecraft on the 12th operational cargo delivery mission to the International Space Station.
<bofh>
today in "ugh": can people please please PLEASE stop doing linear least squares via (X^{T}X)^{-1}X^{T}y?!?
<bofh>
computing (X^{T}X) is often numerically unstable, (X^{T}X)^{-1} is often both numerically unstable *AND* slow.
<bofh>
use either QR decomposition or (especially if lazy), SVD instead.
<bofh>
this brought to you by more buggy numerics code I'm exasperatingly replacing with calls to LAPACK.
<egg|work|egg>
bofh: at least they're not computing a determinant?
<bofh>
egg|work|egg: I've once seen SVD computed by (and I swear I am not making this up) doing an eigendecomposition then taking square roots of the eigenvalues then artisanally assembling together U and V.
<egg|work|egg>
. . .
<egg|work|egg>
. .
<egg|work|egg>
.
<bofh>
condition number: might as well just say DBL_INF (I have no clue what it actually is, but the error in it was large enough on *nice* input matrices that, uh, yeah).
<bofh>
I don't even understand why people roll their own basic matrix algorithms. I mean EISPACK has existed since basically the early Iron age.
<egg|work|egg>
bofh: yes, though I disapprove of using math.h conventions to refer to infinity
<bofh>
and LAPACK since the Bronze age
<egg|work|egg>
bofh: oh, well, writing one's own numerics libs can be fun
<egg|work|egg>
but please know about numerics if you do that
<bofh>
would you prefer a lemniscate? ∞ :P
<egg|work|egg>
bofh: much better, esp. since the condition number is independent of floating point considerations
<egg|work|egg>
if you needed a binary64 infinity, just say so though :-p
<bofh>
yeah, it really is. and I do it for special functions a bunch, but fuck it's so much work even getting down to 1 or 2 ULP error max over the entire floating point range.
<bofh>
(someday I'll have correctly rounded everywhere J_{\nu}(x)... someday)
<bofh>
(irritatingly the problem is the same as always: cancellation near zeroes)
<egg|work|egg>
bofh: correctly, I think not; faithfully perhaps
<bofh>
okay, faithfully. yes. bad wording, apologies.
<egg|work|egg>
<egg|pedantic|egg> = 1
<bofh>
also had a fun incident a few days ago when I was getting unusually bad error around zeroes for Ai(x), then I remembered that I reused my testing harness from some random audio code.
<bofh>
...which means it had the following code snippet in it: unsigned int a = 0x9f80; __asm__ __volatile__ ("ldmxcsr %0" : "+m" (a));
<bofh>
hint: default MXCSR is 0x1f80
<bofh>
(also I should probably make that 0x9fc0, since I don't care about early P4s :P)
<egg|work|egg>
flushing denormals?
<egg|work|egg>
are they called denormal or subnormal anyway
<egg|work|egg>
or denormalized
<egg|work|egg>
or paranormalized
<bofh>
yep
<bofh>
turns out if you have FTZ on and your function evaluation is susceptible to underflow and you aren't expecting it, your ULP error is higher than you expect too.
<bofh>
that was a fun one to debug
TechnicallySleeping has quit [Quit: Leaving]
<Qboid>
0d 0h 30m 0s left to event #1: Falcon 9 • SpaceX CRS 12 [at 2017-08-14 16:31:00]. Say '!kountdown 1' for details
<kmath>
YouTube - Max Canada Lynx - I'm a Big Baby 2
<egg>
kitty!
<soundnfury>
!wpn egg
* Qboid
gives egg a Saturnian DWARF
<egg>
!wpn Fiora
* Qboid
gives Fiora a bow fluorinated guillotine
<egg>
!wpn bofh
* Qboid
gives bofh an isentropic pen holder
<soundnfury>
isentropic with respect to what macrovariables?
<egg>
bofh: I'm confused by ляпунов eggsponents
<soundnfury>
(something something second law trickery something Jaynes)
Snoozee has quit [Ping timeout: 186 seconds]
<bofh>
egg: what about 'em specifically?
<egg>
bofh: exponentially more confused as time goes on :-p
<egg>
bofh: so they're dependent on the chosen initial state?
<egg>
bofh: but are they a constant of an orbit? because they seem to be calculated by integrating the orbit (+variational equations)
<bofh>
yes, they are dependent on initial conditions
Technicalfool_ has joined #kspacademia
<bofh>
well, can be, at least.
<bofh>
(for instance for an ergodic system or component, almost all starting points give the same set of exponents)
<bofh>
so if the orbit is in isolation and unperturbed, then yes it intuitively should be constant of that given orbit since the phase space is the same
Technicalfool has quit [Ping timeout: 383 seconds]
<egg>
bofh: so say I'm taking the evolution of a Hamiltonian system, the eggsponent will depend on the initial configuration (duh), but would remain constant along its evolution?
<egg>
I'm not sure I really understand the implications
<egg>
also, there are going to be *a lot* of Principia L-versions
Wetmelon has joined #kspacademia
* NonlocalRaptor
? UmbralRaptor in the bad poster preparation. With a ?.
<egg>
stabby bears
<NonlocalRaptor>
Poster lists my affiliation as Missouri State University, rather than George Mason University.
<NonlocalRaptor>
Poster also mispells San Francisco >_>
<NonlocalRaptor>
(Is it bad to spell it as "San Fransisco" in 3 different versions of the poster?)
<bofh>
egg: the exponents are a property of the dynamic-ness of the system's phase space, so yes the *exponent spectrum* will remain constant for any given path the system takes thru phase space :P
<egg>
yay
<egg>
!wpn bofh
* Qboid
gives bofh an unitary buzzard/zygomorphism hybrid
<bofh>
> also, there are going to be *a lot* of Principia L-versions
<bofh>
iirc Fourier's thesis defense was staffed by Laplace, Lagrange & Legendre.
<NonlocalRaptor>
?
<egg>
bofh: a funny thing with these 19th/early 20th century french papers in the way they refer to their peers, as "Mr. Lastname" (e.g. "M. Fubini" in that example), rather than the modern academic "Lastname" (or the 18th century "the great Euler")
<egg>
bofh: yup, saw that, that's why I'm saying this: the author is listed as F. Lastname, but refers to peers as "Mr." (Lebesgue, Fubini, etc.)
<soundnfury>
egg: Wealth of Nations is like that, lot of "as M. Lastname has so thoroughly documented, <something obscure about corn prices under the 23rd of 4. Henry II>"
<bofh>
egg: huh. oh, I see. interesting.
<egg>
bofh: and the Gallica example is M. everywhere
<egg>
(it's a summary by Borel)
<egg>
bofh: hm, further question: how long must one typically integrate the variational equations (with periodic renormalization of the perturbation) to get a reasonable estimate of the ляпунов time?
<bofh>
depends on the system I think, not sure how long it is for a typical orbital analysis. usually you have sample data that you are working with, not analytic expressions, at least whenever I did it. that being said, there's almost certainly a paper somewhere on this.