GeoSci.xyz SimPEG, 23 Jan 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SimPEG meeting - January 23

Description

No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).

A

A

B

And so there's no big project, yeah right yeah. So let's go with that and then there's a couple: pull request, updates I, think there's two that are really close to ready to go in so that discretize three much sell grads and did you see all requests so just like a couple things to chat.

A

C

So yeah you're right this comes from that every West one, our most, our strongest supporter yep yep, whatever he was doing, don't work in the US and he's doing mostly grabbin that, oh and then it's problem was that they have a very good understanding of what's happening in the basement was good at sizing, so they have very good understanding of the big structures because of the of the area, and so he would like to be able to obviously recover that was in the conversion, but he doesn't know whether the values should be or the values of each units should be in terms of density.

C

So that's one that was the first problem. So how do you invert for big units at the same time? It's unit one two and three, even a density, XY, XY and Leslie asked for four, though, and the first fast mr. King can we convert with us. This is homogeneous, homogeneous map that hopefully they learned basically to Namib in all the cells in conversion model into quite a few arguments.

C

The other issue is that this is a synthetic model where they kind of like starts. This problem is that if you look at the unit they'd, that map is that he knows about the big units you had one to m3, but there's other stuff. mmm So if you try to use straight up from Virginia's inversion and there's like high frequency content than your data, it's going to use Siri and I. You trouble right to invert to the three units. Converge was going to try to fill it all sort of small.

C

The smallest features limit, yeah, okay, what if we just lay our invariance, who could have both homogeneous inversions and, at the same time I'd like for it originally I, took an inverse of all the units and then whatever extra stuff. That isn't is a that. The same time, that's Yeti other than having a sum summation.

C

They've been assigning two components: gonna sum the three units, plus whatever extra additional material errors. Those are the two maps and like. If we do that, then we can do pretty good job at everything from our one two and three Plus divisional additional blocks.

C

D

C

D

Coal, coal, man and I change.

D

My all the structures are thanks and then you're just basically diverted for three rounds about three gallons. Then after it like Demi using some map, there lurking objects together, 9 7, 3 plus so.

C

Your forward is a sub model.

D

Spaces they can be trying to like put my dick is like it's kind of similar some ideas, but he can't started with parametric version and through these data and starting at all, then you put all of them together, but you could see your choices here.

B

D

Just do second inversion like it is it yeah.

C

I mean in my mind, this is this is better if it's between say a boxer you're not going to fighting for two possible solutions and the difference between this I'm just doing our fence model in Bergen reference model is that's here. We don't need to know. The values of people needed are all to be true values.

C

They just indicators that there's an.

E

Invalid no problem.

B

Yes, founding this, some funding is done with my partner because they're.

B

C

But no that's here, I will need you have a smaller projection. Smart projections take into account that there was some not yes, because then we do this. The projection on the summation and then bookmark renders would have to be forced or we could be the critic another one out.

C

So right now this is been speak, so it's not a problem.

F

You might have like the hold on.

D

Please, as far as the.

D

Property, but that property value.

C

Katelyn something actually yeah that makes more sense again and while you're talking India is let's see. Let's do we were doing strictly positive physical property. We only get other some that there's some of the truth is balancing your who may be. We could have no genius value that is high and the authority makes of lower.

B

Its order of operations like for conductivity conductivity just make sure the wall goes on the website of the some yeah and then.

C

Yeah, maybe that's to get some code.

C

That's I put a notebook.

C

And it's what your figures, so this is I might actually. This is a Mac problem.

C

So I'm doing in the first cell is it's great thing. If that's sort of a and if I, minus twenty to twenty meters, that all my receivers and then I'm creating a model with just it's a black. It's a black and a half space about the 0.05 aside and then I'm going to just going to split the domain into two into two parts. One has zero and yet, on this point, zero one. So the model looks like this because the plan view it's the cross-section in juice.

C

It's just a block and we have unit one in immunity.

C

We can run the forward elimination that looks like this so because, like I said they tryna coming from coming from the east rate, it's going to harder to see, but they give you the blackness in the center. Are you, but because there's such a big trend, that's kind of over over printing and everything?

C

Okay? First, we can run a traditional, smooth encouragement and here we're in both we're just saying. That's me: zero, as opposed to your wet spot so everywhere, so how's, the I'm getting two buildings right.

C

The block is sort of the right place, but yeah it's the original. Every hard time is because that's something that's wrong and groceries wise, there's, actually a big tram.

B

Doing here and then are you using that's, not that's. It's.

C

On the Cape so.

C

Okay, oh yeah, you see the first experiment. Well, if we just what, if you just add twice the same model parameters, we're just going to the model to two of the same all right, so each doesn't have to be the same, so hopefully I just give them all of us have the size of the other one right, because if we just have.

A

As maybe, please go ahead and.

B

Just we, why don't just say what you're gonna do for memorization? It's both the background.

C

C

So the first one I give up we're not even going to do a homogeneous inversion right now. I've just said it when I show this one here, where we have one one: Victorville founders, that that's number of cells in the mesh.

C

I'm, just gonna grab this guy twice.

C

To my tour operator, so basically I'm I'm, saying use it twice. So hopefully, the recovery model using me have to have the values that I'm getting here, because I'm adding it to itself twice and to the four twice, and it's like what what happened? Why do you look at the key values from this.

C

0.008 and this one is point O 16 right, so I regret over model. It just happened itself because I ended it twice: two inversions just a business if, as I said, everything.

C

Now what I will do is.

C

That wasn't so all I did is the Sun mouth and I just put two identity mapping, so I'm stacking twice.

C

So that one round.

C

Not what I can do is is create split. My boobs good, my model space in the manatee domains model a model B that I know. I know this exists.

C

So I'm going to create animal Tina smell what the emergent is.

C

I didn't put a picture here to explain, but basically we'll it knows which model parameters that vectors exists or belongs in Group one and the front up to 52. It's not a single value, each of these groups and then, when I, do my forward model I, simply sum up all those cells times that value.

C

That's what is.

C

Now I can join the homogenious like this. Don't worry too much about I'm going to stack and there Oh genius. That means all the cells are active. All the same selling size number of it. I'm gonna, have my combine.

C

And just add this to be to the for decoration, so it knows which model parameters are due to do for the routine and which model driver's.

E

License cells, prison.

C

What's your accessorize, yes, is this this disc right here so my my only genius are two three whatever and that's stack on top of your fold and see number of cells, and then the Sun map notes take a projection of first, you expend it to full and an additive to the full vector I said it. Does the summation of the.

C

That was four level inputs browser.

C

So we also need wires all right. We need wires because that's what those lines are. Those are the wires the wires, no knows which one's belong, to which a.

B

Model vector and basically saying these four parameters are actually associated, so in Indiana you could say your model for people in the end, but then you need to tell me emerging like these eight factors. Actually, these are.

E

B

And so, while there is just a helpful way that maybe.

C

So but the name wires.

C

Yeah really went there wires, it's just it's just about matrix number of parameters at fullback there, and then it collapses to whichever who wants to grab from that from that dangerous. That's what it was kind of that interest. It is that the things yeah.

C

Okay and so of a sunbath I'm doing it on a bunch of here do a little bit of a scaling for the distance wave thing. Sento are since these the Tigers weight. Then you can figure it out. That works pretty well, and that's even your regularization.

B

C

Yes, all right now, I'm just continuing it back there a year. I am just collapsing the rows of distance. To do this, I am doing the opposite. Scaling for the emotion you smell is that I want to preserve the fact that bigger units have a higher sense, that on a kill, kill that part so I'm just going to divide it for now. So when I multiply that's good and then for the origin, neatly it's just the same as usual. It's just some rows of the sensitiveness.

C

Now the realization not asked to pass two parts right. We can have a part that it's gonna recognize the homogeneous part and the cardio is going to arise in this part, I'm going to create two organizations and I get me get that locked together, because we won't know who you might want to do something that and a notion for examples, for instance on the authority.

C

If you want maybe want to smooth- or maybe you want to come back, so we're going to be able to swish you can assign whatever you want either acid yeah, that's whatever just last name. So this is one example that what the Sun map is doing. But now it's opened up to ok,.

C

So first do that l2 and then we recover this this model, so we go value could uniformly on this black.

C

These there's an issue right now that I still need to figure out. Is that because we've only summing, one side is going to have higher center one, so I need to think in terms of regularization of organizing this, both rather than one one time water. Initially, things like open.

D

C

A ensure each one, just yeah, just the zero genius.

C

All right so just the authority nurse fire, that's that's the now we're adding this to a model that has gumming eight, which is higher than two so we're just keeping up once I deprived is the summation is going to be higher.

C

Which my mind met beat? Which mine is not good.

E

Hurt right on to go greedy.

E

Rafinha studies, it areas outside of that table background values, mm-hmm.

C

That of your father, yeah, that's good! That's good wine! I'm in this part, so we're doing a good job. If we ask for that, but not losing it kind of like they like help, somebody still have a bit higher on this.

B

C

Different gonna be well. The thing is that the suite model extends extends their rights, so it will sort of add both. That means just what is the yeah if I do a wire line.

C

Those are the two values so point, oh one small, so this one is actually good right. Our issue are true models that there's a little bit extra.

C

We could we could do sparsity on the background on the background values as well right.

C

D

C

That because they get pushing the other doubt exam, let's try to program this or if you are large enough.

C

So yeah! Well that runs this. That's these maps, I kind of like to be served up some issues right in the way, the derivatives, what go Enzo I mean because I mitch is what I'm worrying glad that my working I realized.

B

C

The in the right, I'm, sorry yeah, both in of some out and problem. We have right, yeah so know the purpose of the mapping.

C

B

Question I have just one within between doing wonderful, so some advocates like each district who I don't do you want to call it something like I do.

C

B

Used sir James to indicate when we're going from one value to do themselves, don't look there's mr. J old. If you want to know something, they've got.

C

Hey: what's this, you can embark on that I producer Jeff, so.

B

All that so we kind of like circuit and eject right so index is 1 1. So when you're doing it ourselves, for example, for mapping like exactly the number of selves now, just so there's no change in the number of words between phonetic, whereas when you surjection you're changing the you know for two grounders to now themselves.

C

Man, if I believe nobody else is doing this happen, yeah.

B

Well, we have is with Serge X well, we don't actually have any other style or surjection, so we only have one that goes from one.

D

For mommy to be a virgin.

D

B

That don't you did yeah so they're also like this to more classic.

C

So I create a PR.

C

And then yeah, it's not ready for it. It's not really really very humans to surprise for our something that's worth and then go I'm sorry come on. Have you come up to you.

C

Yeah so oh we're gonna miss our clients.

C

What's the first motivational example, it's this it's the Porsche, the Porsche that said I was talking initiative. So it's not it's not exactly.

C

It's an innocent that looks very similar to this one.

D

Victoria is actually on the bomb range. It's on the money, basically you're kind of isolated target song thanks.

C

Like it could be anywhere good young Bob, it's just like it is just a problem where you have a short way of life, Sonny's broad background, so you want to be able to birth through the chin and.

C

There's one other time: the original are gone. If you don't back without use of your dog, your big, the wings, is going to be wrong for sure if you're, including this enough and then still have our land conversion is going to try to do this.

D

If you do with your key way that what nice thing about you, don't really worry about, that, it's graphically miss Charlotte, turn boundary and he and the third final and say the second duty logically a king can it is always doing Paul. So.

C

That's true, but I get so you would use it a favor, that's novel! Every time you cluster in between now. How do you go from I?

C

Think it I couldn't give a para mentary when you point you in your primary conversion phase or something somehow needs you everything: okay, yeah, that's true!.

D

Yes, I see your point look like if you were doing what I was curious because, like that, how bad it is, if you just cannot hold things like doing, every single sin then just do whole it's to the first path, in version of oh geez, man and I. Think.

D

That is really.

C

There's always going to be like in your circus, and ah these are just shrunk them back there. The thing is to actually interested them. You're not going to be able to fit up cuz. Your your job model doesn't happen, but they've all the details right and then, and then you can say I gave all my head, my inversions style because I think video. But how do you even know that what you wanted I stalled outside stuff? This is the best I can do yeah.

C

Well, when did you could exactly I mean you do I, don't know it sounds like this is such a piece here is more streamlined, yeah, it's much easier.

D

It's much easier I would hope that again cuz I, think of all the fun dimensional weight and say who can think about as a regional me, you kind of carve up as we get that Matt I got two of novel ways and I have an idea. You can just carve out some portion of that and then do a whole. She gets like like I'm using that Matt go ahead, version cover background and then I can have those please now and then just yeah.

C

That's a lot of steps in and we are does not that it needs to be like if you're, if you're a security short way of life is like that somebody bizarre shapes, like you, don't even know what that footprint, that that is our shaping status, dude and add the whole thing with regional removal is that the identity we tried to just invert it the battery room, the tram, then we try to with Jeff I did you're my in Spain. Initially, when I would look at it. I said like this I would say.

C

While there was like there left to right, there's an east-west trend. I'm just gonna have to debate that an inversion major with that is that there's actually signal. There's interesting signal that y'all just want to know they want to know where the basement is, for those dipping features are, if you just blend it, remover to polynomial trend. You know, yes, you're gonna be able to recover the small features, but this is the whole story. Is.

C

They would they would get there two targets and then a new record yeah Lori just leads to more questions and yeah. When you remove trance to the dangers that it's a bibbing, you know you might be killing a so everything's gonna. Have that, like you straight vertical.

C

And it's the first seven, it's just a mechanic's brother Jeff.

C

Yeah pretty passive yeah two weeks of work and now it's running.

B

Yes, so you guys want to like briefly chat through justice. You major for quests that are on right now,.

B

B

There is a big change to me in probably this week, so I'm gonna briefly just give an overview of what's going on with this PR and what.

C

You see the status, we're almost passing our newest webinar coverage. It was msn.

B

Yes, sir, what don have been working here for Tyrion? Yes,.

B

It's been working on the celery needs for the operating physics. You sort of take a naive approach. What ends up happening is you get large, here's your little weird cell sizes and so yeah we're we're it organs so because, if you use like the native approaches to boots the transpose and that doesn't account for the pack fit like all of these four cells.

C

Ended up with a longer radiance fits so you're over smooth in every time. You change effects. You devil yeah. So now I'd like we're pretty good right like there, so the top was before Vienna at the bottom. So it was a lot more like whether a dancer inversion. Obviously the online course ourselves. But you don't have you don't see you as much.

C

B

C

Texting, that's quite a bit yeah yeah on that PR. There is now the public gradient cifaretto's. That's.

C

What did you end up doing Hannah Ritter passed through it, but he took me self-paced, given them decided. That mention was something to him. Did a couple different things now: I'm.

E

Not sure they're 2015 he later.

E

On? For the nerves of the testing stuff, to be able to.

D

Know I think I know what you think I'm not at take holy get into our mesh. It's again so there's a lot of Florida thinking of comment, the volume and then there's one part are they were disgusted with my authority think to do what economy without yeah, that's right, so I think Mike and I was working on with Foley. So I wasn't following what dump did I post using as a simple trick for a big corporation and basically using a given festive and then just applying to the apartment? It actually get that equivalent.

B

Because one krishna had the so if they tested on a uniform three months, but not on a branded thing.

E

D

That, if then, we may not get out of second order accuracy: okay, yeah, not the hundred percent sure which tested it. It's basically following together together test yeah.

B

Donning not three measures.

D

I wasn't really sure what to expect it's mine, my guess, is supported I'm longing for himself at rematch, great spectacle. First of all, practice that would that was funny. Yes, yeah.

B

That's when we get the guy. Thank you because I.

D

Think it should be read, it should be secular directly right that there is Kabul things kiss, even though, like there's a good one, it's not like I'm gonna publicly with the Pesce, actually degenerates a circle and then refine in the circle, and so it's not quite there will be.

B

C

That's probably.

C

Something we could do desisted splits yeah, ideally we'd like to be able to jump just bus.

C

E

B

This, the Indy serious it actually.

C

Don't that's in boobs, okay, let's say saving models, sir, but I thought maybe take me dead. Yes, yes, yes, yes,.

B

C

This is thin: oh sorry, yeah you're, right.

C

B

Are you gonna give a quick update on.

B

D

B

A

Wanted to show you start off on careers. Just yeah, okay, yeah and I need.

C

B

Well, we're so online.

B

We started the blog again.

B