Tumgik
#merlise
trashcreatyre · 2 months
Text
Tumblr media
The uhhh
Idk i just love Merlina and Elise :3
260 notes · View notes
leighmuold · 5 years
Conversation
TEXTS || MERLISE
MERCY: HAPPY BIRTHDAY LITTLE SISTER!!!
ELISE: !!!!!!!!!!!!!!! THANKS BIG SIS
ELISE: Now I want a dramatic retelling of my birth, including the part where I almost ate Kristy in the womb
3 notes · View notes
teamxdark · 3 years
Note
From the sunset horizon , after finally fixing my sleeping pattern so I'm awake during daylight, and buying bugsnax, I return with a very importent tomo life update: merlina and Elise are now a couple, Lancelot and Amy are now friends, I awkwardly laughed as mighty and tikal got married, percival apparently really hates milk, and most importantly, everyone say "Thank You Zephyr", as sonadows son set Arthur and lance on a successful date, and now Arthurlot is canon
Tomo Anon! Welcome back!
Merlise is really happening, huh? Excellent, good, these girls fantastic.
Lance and Amy friendship!!! Lance and Amy friends!!!!!!
...WELP. Ah, it's fine. Alternate timelines and all that. I know they're not canonically siblings and even though my own canon affects the way I see things, I'm not gonna be that guy who forgets that their headcanons aren't canon. (Especially for an AU of a semi-canon AU.)
Percival: If I have to break every stereotype in existence then that's what I'll do.
Thank you Zephyr!!!! Finally!!! The sonadow boy just wanted to see his alternate universe dads be happy!
1 note · View note
notstatschat · 7 years
Text
Fixing an infelicity in ‘leaps’
The ‘leaps’ package for R is ancient -- this is its tenth twentieth year on CRAN.  It uses old Fortran code by the Australian computational statistician Alan Miller. The Fortran 90 versions are on the web, but Fortran 90 compilation with R wasn’t portable back then, so I used the older Fortran 77 version. 
The main point back in 1997 was to provide a version of the ‘leaps()’ function in S, which uses a branch-and-bound algorithm to do exhaustive search for the best (smallest residual-sum-of-squares) model of each size.  I was interested in exhaustive search back then as a step towards multi-model inference -- either formal modelling averaging, as Adrian Raftery had taught us in BIOST572, or visualisation of what variables were substitutes or complements for each other.  For example, did systolic and diastolic blood pressure tend to be in together (so maybe their difference, pulse pressure, was important), or did they tend to substitute for each other. The `plot` method in the package lets you look at this sort of thing, inspired by a talk I saw Merlise Clyde give.
However, exhaustive search isn’t as popular today -- despite the clever algorithm, it takes a long time. It has to; it’s an NP-complete problem.  People have started using the forward and backward search algorithms in the package. I included them mostly because they were there in the Fortran code.  You’d think that would all be straightforward, but it resulted in what is either a bug in the code, or a bug in the documentation and a feature in the code that’s less useful than it was.  
If you look at efficient construction of an exhaustive-search algorithm, the working step is rotations to swap variables in and out of the model.  It’s convenient to start off with a preprocessing step that fits a model with the first variable, a model with the first two variables, a model with the first three variables, and so on.  Everything ends up getting initialised; it’s basically free; you then go on to do the search. 
If you’re doing exhaustive search, those models would eventually be considered anyway. If you’re doing forward or backward selection they might not be. However, if you think of forward or backward selection just as a computationally-cheap approximation to exhaustive search, since you’ve got those models anyway, you might as well look at them.  That’s what the code does. It’s possible for, say, the model with the first three variables to be better than the three-variable model found by forward selection. If it is, you get the better model. 
In the modern world we don’t really think of forward and backward selection that way. They’re specific fitting algorithms you might use as part of a selection/regularisation strategy. They’re also things you might teach for historical reasons.  In either case, it may matter that the sequence of models you get out is nested; you don’t want a better model, even for free.
So. Bug reports. Attempts to debug twenty-year old Fortran code written by someone else. Attempts to tweak twenty-year old Fortran code written by someone else. Searches for work-arounds. Swearing. Finally,  version 3.0 of the leaps package, now on CRAN. 
If you have nbest=1 (asking for just one model of each size), they will be nested properly for forward or backward selection.  If you have nbest>1 I can’t see how to avoid the extra models, but you will be warned.  As always, this isn’t an issue for exhaustive search.
2 notes · View notes
ao3feed-merlin · 4 years
Text
English Courses
read it on the AO3 at https://ift.tt/2JG6GR2
by merlisation
Short story about someone who will be important for two boys not yet in love. But it's her introduction. Not much happening, maybe too much dialogue.
Words: 882, Chapters: 1/1, Language: English
Fandoms: Merlin (TV), Original Work
Rating: General Audiences
Warnings: Creator Chose Not To Use Archive Warnings
Categories: F/F
Characters: Original Female Character(s)
read it on the AO3 at https://ift.tt/2JG6GR2
0 notes
trashcreatyre · 3 months
Text
Tumblr media
If I had a nickel for every time I’ve drawn Elise braiding her lesbian partners hair I’d have two nickels
X
49 notes · View notes
trashcreatyre · 23 days
Text
Tumblr media
Hi………what if……….ponies…….?
38 notes · View notes