GeoSci.xyz GWB Lectures, 20 Feb 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2020 02 20 Day9 Part3 Regularization DOI

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

Roll it okay, so we're back so I know there is a lot of complicated things that we are learning in this class and it is tough to remember how to do everything so need.

B

Not have a dime, yaaay, you've gone home and don't need time hope. There's some video by all messy I'm uh me neither but I think about it and.

A

So getting the Jupiter notebooks on your computers is a challenge.

B

Yeah hey da day and I been kicked it out of it. There.

A

Are a lot of things you can do with process to that? It would be nice to have some documentation.

A

And to load the data and get to an inversion has many complicated steps.

B

Ambassador but I really I should wait a minute. I said: yeah, yeah, okay, so.

A

For each of these things, I'm going to make a video that will just let you follow the same workflow every time.

B

I mean mass or guano I think you know how about it. The same is: leave the do these.

A

Are YouTube videos, youtube I and the link to the YouTube video will be on the Facebook web page Facebook.

B

Page my head out about the NIU to me so.

A

We don't have to go through the entire video, but maybe we can just go through the first five minutes.

B

A

This is a 17 minute long, video suck when.

B

It's out, if you do well to.

A

Get you having the Jupiter notebooks on your computer for.

B

Me that oh yeah yeah, we do be some kind of Miss.

A

Correctly, you will be able to use the notebooks without meeting the internet.

A

So the first step is to install anaconda, so we will show you where to download anaconda and how to install and the links for this are on the Facebook page via make Myanmar hydrogen visit society.

A

So there is a link here, thank you to download anaconda, so we will click the link. It should take you to a page that looks like this. If the link does not work for some reason, then you can type download and a Conda to search, and it should be the first link which shows up.

A

You can download and it will take you to the same page. So if the link does not work, you can even search. We see that there aren't two options.

A

We want 3.7 version as you can download, 64-bit or 32-bit, so you need a different version for different computers, and so we want to check this so to check which version you should use on your computer. We will search our computer and you can type the words operating system as operating system, and this will tell us about our RTC, our computer.

A

We may also 64 min or 32.

A

This would be another way to search, and it will take you to this page.

A

So we'll read some of the things we look for system type and my computer is 64-bit operating system, so system type, 64-bit operating system, so I will download the 64-bit version.

A

So this will take a few minutes.

A

Okay, so we can see in the bottom corner the download has just completed this took about ten minutes, so I might pause the video while it was downloading and we can click this button here to start the Installer program.

A

You can also find this in your your downloads, folder, so I could look in my my downloads, folder and I can see. The installer is online.

A

So I think next week and I will choose to install for all users. This just means that if more than one person has a password for this computer, everyone can use the.

A

So I will agree that it then download it and now I need to choose a place for the program to go and the computer is smart and it will pick a good place, usually C program data and a quantum.

A

Computer is me, and you click Next. It will start to download, but if you get an error or a problem like this, that says this, this place is not empty. It is because you have tried to install anaconda before so. We need to delete all of the previous anaconda. You have on your computer, so then we need down to be installed. We have one version.

A

So one trick that you can use to find this water, if you have things inside, is to select lights, click and coffee.

A

You open up your computer's folder any folder, and you like this and you right click and you paste and if I press ENTER I go to the place that I paste it.

A

So we can see there's only one entry here so.

C

We need to get rid of these.

A

So I think you get the idea of the video, so this will go through the rest of the process to get everything set up, but you see you can just follow along with the things I'm thinking.

B

See like they have the same years ago, but if possible every possible you know about it and.

A

So tonight, or early tomorrow, I'm going to try to make another one with the process too and I'm going to try and to make one where you have the data file. You load it into the notebook and you get an inversion of a cell for.

B

Another group that well table a poor not resist as well and.

A

And when I finish, I will put the link to the video on the Facebook page. If.

B

You I know and make the block is a base, but then Damon am I going the right way to do that me so.

A

I'm making this for you, so if you have any suggestions or if there are any things that you want, you could let me know now.

B

Look at the Mohawk, the unrequested entirely the other way in it. A lieutenant Bates and I found ability. Why bother I'm announcer? It's a level! The illegitimate, my terrifying night, every grow.

A

So you, you are all happy with this. That's Indian I'd, like.

B

To go yeah, okay,.

A

B

I'll just keep.

A

Doing it the same way that.

B

As a young yeah I'm, looking down into like a little salad bar menorah.

A

Cocoa can we turn off the volume.

D

A

Adjust it'll just be fine.

B

Or giving them first.

B

D

B

A

E

Is anybody using masti DRD because max couldn't get it I I just lost my caption.

B

Using down than the White House spot.

E

D

Well, I think I can.

E

We could probably get, but so again how many people were able to run the inversion with my binder everyone who could not. Who was not able to run it. With my binder my mind on every.

B

I know she got it might not. Does the embossing putting.

E

Everybody perfect perfectly okay, so I will now just do a couple more in inversions to show you about depth and investigation. So I'll just do it locally on my computer and you can you can try to follow along I'm yeah I'm. Just on my can I'm offline just doing.

B

It internally so doc the ended up, seeing it oh.

B

E

I think the main thing is that they've been able to run one inversion, even just using my binder and so they'll become familiar with the with all of the middl parameters and patents that are associated. He.

B

Must machine yet the body healthy? Our example will not. You got me down in class when I party long ago to this one label about.

C

E

I'm going to vote again the block to the dipole-dipole noisy data, a mother.

B

Da-Da-Da-Da-Da-Da Bobby the deal if I receive it.

E

And I'm going to plot the data, there's.

B

A tacit assumption, that's a position without me,.

E

Get a mesh I'm going to change the core length, so the core depth length is set at 30 meters here, I'd like to make it a little bit deeper order. Corey ends.

B

Up nicholae well behind.

E

I, wouldn't ever go.

B

Up 10 feet off G's in the bottom. It's an intimate so.

E

I'm going to set the core depth to be 50 or adaptable, and.

B

E

Now they look damaged we're going to have fine cells all the way down here, and then it will start to expand so found.

B

On T cells got me up on my ship. You don't be a cell to do one thing: that I'm.

E

Going to do that for two reasons: one is that we'll get better resolution at death and, secondly, it will allow us to see a little bit more clearly how what's the depth of Investigation of our survey. So.

B

You gotta get in the new she bought it and I'll convince on that drop. An epic. The net would never be denied, laugh.

B

Eight or nine eight eight one.

E

So now the core is extended down to down to here and then it starts to expand.

B

They understand I would what did it do on them?.

E

We've expanded the core, so that means we now have more cells, so we have 5300 cells. Oh.

B

Yeah Stiles sounds.

D

B

Horrible day night, anything about.

E

The set the uncertainty to be the same fighting force at.

B

Bedtime and identity, though.

E

And now we're going to run the inversion result. The.

B

Fashion design, gravity man, I'm.

E

Going to change this 2.0 whoops point: oh one.

B

E

Alpha s was that parameter that was in front of the smallest model component, smallest one, the smallest moment, member.

B

Excope, no more a diesel.

E

Don't you remember, the model objective function was alpha s times confirm that made it close to a reference model.

B

Or a stop, it represents more than enough plus.

E

Now found X then smooth in the X Direction plus.

E

Alpha zet smooth in the Z Direction saturation.

B

It's an acceleration receptors right.

E

So the objective function was composed of three parts want to try to make a final result. That's close to a reference, one that tried to make it smooth in the x-direction and one that makes it smooth in the set direction.

B

So if I don't wish it I, will they see they're happy they're cool ah the young man smooth syrup, we're gonna smooth only each other I looked at it like cool I, might eventually marry that hot girl. So.

E

We explored these on Monday if I make alpha said really big of, as everybody doesn't know, terracotta then, what's the character of my model what's way.

B

Won't have a lot of money of Dalmanutha alpha executive gala, so.

E

This is really big, then what what does my final result tend to look like one.

D

Of those I'm into attack.

C

E

Magically smooth fur and spread so it's dead smooth in that vertical direction and if I make Alpha X really big, that what is what happens as.

B

Smoke horizontally horizontally, smooth and.

E

If I make alpha s really big, what does it try to do that watch.

E

Yeah it'll try to make it close to who's a reference one. So.

B

If, if you guys Borodino now more than ever, it will be said if.

E

The reference model was a hundred old meters that it will try to make the model a hundred meters, so.

B

Our reference Maria yeah, oh, he doesn't miss a any cacao, say how.

C

B

Hundred would be gone, but let me go mini salad, but.

D

E

Course it can do that because the data would say a hundred ohm meter. Half-Space does not fit the data.

B

I'm space modality it does that automatically dim its average, but.

E

The the program, the inversion will try to make it close.

B

To them whether the application out of them is out of 100 automatically yeah, okay,.

E

So what we're going to do is just to run this. This model.

E

So we've run this before, but we've changed now the mesh just a little bit. We have more cells.

B

Sauron be lobbied, I met Juan, Manuel chained me to a corn meeting when I got up. They pulled ourselves in the woman, so.

E

Now, let's take a look at the date, that's the data and the model.

E

So here's again here's our data misfit with iteration.

B

So we in duration and animistic jamison I'm gonna we've got our characteristic.

E

Curve and our model mark keeps increasing, so everything looks fine, yeah anything and if we look at our data, misfit.

E

Again, that's just the first iteration. If.

B

We look at the end.

E

Result, that's.

B

E

We see we've done a pretty good job of this.

B

E

If you notice in here, let's look at the model, so here's our model.

B

E

We had tried if we look back here. Our reference model was eleven hundred and thirty-two, so roll wrap is equal to eleven hundred and thirty two home.

B

Beaters, that's more than yes, that that block copy other than it only.

E

Down if we look at the resultant model that we've got here, we have more now.

B

How to use it, we.

E

See that a lot of these areas have a reason stiffening of about eleven hundred old meters in.

B

Mariana who reference morning above the neck, I've every Monday, so this beyond around the rushing yeah yeah.

E

So we don't really know if this 1100, all meters out here, is caused because of information in the data or whether it's just simply a matter that we had put 1100 meters in.

B

Our reference model- Eman, sorry, tough, D out so I hope, you're, happy, B, money, moral, yet that hotel yet again settler and the young lady diet, yeah, okay, mom Girardi reference morning want me, get a.

C

B

Socket tonight that will answer on the theta or not so.

E

Let's do an experiment, as I said that today lives yeah, suppose we change the value of the reference model.

B

Reference water, yet that voltage changes so.

E

Row draft, instead of being eleven hundred and thirty, two.

B

Of those than this, huh let's.

E

Make it half that the what I'm? Sorry! So, let's make it! You know 565.

E

Okay, just pick a number, so what do you think is going to happen in the recovered model? If I now try to have a background, that's 569 Justin.

B

Doesn't not a la talla missing anything become a mother who keep on it yeah.

E

So if I make the reference model lower, what do you think might happen.

C

E

E

So what do you think will happen.

B

Resisted TV Netta, the low resistivity area works, fine paper, okay, anybody.

E

Else got any suggestions. What.

B

Type, are you come on TV today, a table or table so.

E

Let's try that so we could do.

E

This acronym D, oh I, but.

D

B

E

For depth of investigation do.

B

I'm deaf of Investigation: this was that one yet an epi yeah.

E

So here I have a factor of 0.5 that is going to take the previous reference that I had multiply it by 0.5 and then I can run. So that's what we did. We had a previous 1132 multiplied by 0.5, so the new reference model is 565.

E

So I'm not doing anything I'm.

B

Alone except I'm, changing this guy right hand, corner 1110 again,.

E

Everything else stays exactly.

B

The same opponent to retire.

E

So it's finished now: I can plot the results.

E

E

Here's the model that I had before and now that all I've done is change the background resistivity and you can see how much later it is at the bottom.

B

Bottom I doubt it so.

E

For a reason for a half-space resistivity of 1100, we got this for 500 I, get something that looks like this, where it's light here and right here, the.

B

Ones at the town's Owatonna uh-huh how good's I don't know our name so.

E

Any change between these two pictures is telling us that that is just because of my chosen reference model.

B

On top jung-hwan later on, he goes I'm getting about as a reference more that that will tell me there so.

E

I can convert that into a number if I have in places like this up here, where there, the value of the resistivity stays the same. Changing the reference model doesn't alter the the result and for those places, I'd say, I've got good depth resolution Beaumont.

B

Ayana McDowell walk away all right: Nina, hi, Natalie I went down garden godunov, others are so ludendi, but until an epic ramayana or not for.

E

Faces down here, where the value of the inversion result becomes close to what the reference model is, what I've changed the reference model that say I have the poor resolution.

B

E

My end result comes out to be it's what I put in for a reference one. So if this result down here comes up to be 565, which is just the value of the reference model, and this comes up to be 1100, which was the value of the reference model, then I know that this pixel down here is the influence only by the reference model. So.

B

B

Any reference model when we kept roto bar below, like it like.

C

You don't need it as a reference for analysis of a person, so Takagi reference.

B

Morin a little reference for a chain: okay, we have people that make what you are.

A

Quite right as well yeah.

E

So there's yeah, so there's an index that goes from zero to one if the value of that index is close to one which is close to zero, sorry, which is which is the red. That means that I've changed the reference model but I'm not changing. My end result ya.

B

Know some local young man, a god nodes, not neon feet or zero to one I love. I love updates how much a reference for each English Aguada, but.

E

If I change my reference model, my inversion goes to the that state result of whatever the reference model is. Then the value of the index is one.

B

Index it neva take Virginia Lomonosov, Ojinaga DW. She she she she, the main reference for you that would be on I got a reference model named Eduardo I mean I, might want a matte vanilla. So.

E

Here's a final picture that has got an index of 0.3 and everything above here is thought to be like okay, I'm, seeing information, my survey, it's kind of seeing down to this step and down here, I'm, not really very sensitive,.

B

So there's another boy and the young ones, he'll got a hobby if they make this. How feed you didn't tell me either huh I'll see you at nothing, I'm, never up that you can get out they put on my meow. They put up a do. These aren't happy about it, but.

E

This makes kind of this makes physical sense if I'm sitting here ball. My electrodes are in this region. Here then I'm simply not sensitive to something. That's out here. So.

B

Generate about my net from my mother, she then has way no I'm a poop not sensitive to grow. Hello. We've got it's not telling me I'm us, you know these are about these and certificate very much.

E

So if you think about a BC resistivity survey- and in this case we add two hundred meters- okay,.

B

We see business degree thinking about uh be so.

E

Sue zemanick.

B

E

Hundred meters and man, if you look at what the you know, what the depth of this contour line is, when soul of.

B

Windows on yet it's about like Miss Lee about forty meters, these and you sign up today and.

E

That's pretty close to what max was talking about for a rule of thumb. If you want to detect something, you know, then maybe a fifth or sixth, the electrode buying distance is your depth of investigation. So, let's.

B

Sit down, let's do it. On top of the Baltimore clinical Devon reality I wonder what we animals internet icons on the bottom. You see that your nephew love me nightmares, Rao, yeah poverty do money on a device Topanga.

E

So, of course, you can decide what error index is that you want to to plot. So if you wanted at zero point for, for instance, then that would that would move index.

B

You're going down low Tom! Isn't good luck like y'all a lot unless you're closer to anybody, but.

E

The important point is that there's really kind of very steep change in where you've got indexes of about one versus something that gets close to 2.

E

Zero is you're going from the blue to the red that transition happens fairly close in this.

B

Numb up down nine eight nine now.

C

E

I think that you know something in the order of point three. Your point for seems to be be pretty reasonable.

B

So that the month on with the money that loved ones and they, what attempt you getting about it in.

E

Reality there is no hard cutoff that you, you see above a particular depth and you don't see anything below.

E

If you ask about.

E

So here's your survey length, okay, there's zero, here's 200 meters, then we say that oh, the depth of Investigation, you know, might be something in here. Maybe it's like you know: 40 meters or 50 meters.

E

In reality, there's no such thing as as a line that says. Oh I can see everything here, but I can't see down here.

B

Da da da da da okay, let's.

D

B

Tomorrow, Athena depends on my day cause I'm young enough. So it's so a minute I'm. Unlike the lines, if you don't tell like the lines of the ship, are watching another sub step or today, okay they're, not by Chara yo yo yo yo yo go avec, TOI aussi man cap. You know.

C

Hobert Elswick Allah. We are intimately a numbered.

B

Obama she was robbing my house enough. It's a moronic, your! What is it even worse, so.

E

If there was a boy here down below what you, so, let's suppose you have a block here.

B

And I don't know she did what time I'd like you.

E

Might still have a little bit of sensitivity in the dative to that.

B

I didn't know it our female Oh.

B

Let me net dawn on them. Yeah information data, everything guilty to be on sort of I, see that I make an enemy I'm, like am I. Where are you.

E

But certainly if you had a block in this region in here, you should see it in.

B

The in the alley, one time I, don't see it as a moment.

E

So that is something that can sometimes help you just understand like how deep are we are. We see with the particular.

B

Houser in an idea upward and that luckily an egg layer at Holly, I didn't it and.

E

Here's a formula for.

E

Here is the formula for the depth of investigation index the.

B

Index it well, you know, I, do I index it so.

E

Let's suppose we carry out two inversions one with a grip of first reference model called m1 rep. Anyone.

B

Get set up come on reference model yet hold on, and.

E

Then we get out a solution from universe, m1, the.

B

Solution and well my solution right.

E

So we put in, we have all of our inversion all of the parameters, all the uncertainties everything put in that.

B

Would say what the environment I will shave them easily yeah.

E

So we put in ref one very.

B

E

Ticket which, which would call in here and rep.

B

E

This is like the ref reference resistivity for the first inversion hey morning.

B

In reference water here versus didn, t 1, so.

E

That would be 1132.

B

So help me out those that knit now.

E

We do a separate inversion with.

B

Another for aggression to open it runs loaded, okay and.

E

Then we get a second inversion result.

B

So my mom bought another to get food but I'm in.

E

So if the change in the inversion results matches the change in the reference models, this number becomes 1, and then you know that the data are not influencing the value of that that self.

E

So if, if I change the reference model to be like 1100- and this comes out to be 1100, I've changed it to be 500.

C

And this comes.

E

Up to be 500, then the this number here comes up to be what then come.

B

On Tonto the chain, I get sorry I need morning. My leg I create my I got. What's in there. Where do the audience? You know my eyes. What is an irony buffeted by and a coup go let's say, I think I do yeah. Yes, it.

E

Should be one well, if it's one, then the data are not affecting that an.

B

Optimization or a gigantic data you gotta read in Boston MA I'm, a drummer moonship, a net reference more and I'm already in.

E

Other cases I changed the reference model, but these these numbers stay the same. So in fact, in here.

E

So right in here, for instance, I change the reference model at one time it was 1100 the next time it was 500. These numbers are still just about the same. They haven't changed, so that means that the DOI is equal to zero. So the index of zero means that you've got good lead a good resolution, so.

B

We're not the build on the reference model with Yahweh it don't even matter hey AJ machi de resistivity time warmup Yong, Hwa, Yong Hwa was enough to reach a deal on yet index. Zero Catera, a tsunami about daddy theta, a manga did I forget to give you our reference. More example. Are you resistive? We are not young, you are a representation of a guitarist reference morning for mutti Grover Mika with young or a tdawg. Ahahha did not give ya another zeros are not how many of you are on. Okay,.

E

So does that kind of make make some sense? It's a way. It's a question that people always ask. Are you seeing something at a particular depth, and here is a way of kind of quantifying that Turia.

B

And I think we are there nipping Medina have I do America that I, don't yet see with the Coopertown tiny cut. Erotic manga is wrong.

E

Okay, the other thing that we haven't talked very much about, but is one other parameter to explore, is something called the Chi fact.

B

In fact, if it was about I think that machine Yaya Hyrum, you guys up there like a X, so let.

E

Me, let me just restart my colonel and.

B

I, we start to go like that.

E

I just got to clear everything and then start.

E

So I gotta know the same data staff.

E

Okay, so we're back to the 3600 cells. I'll set, my uncertainty be 5% and.

E

Now we've got a hundred.

E

We have a hundred and twenty four hundred and sixty two data, so the number of data that we have fakest versus one one: six: okay, that's that's the number of the data.

B

E

So generally, if we and we had a five percent error- and that was the a good representation of of the noise hi.

B

I know: I, don't watch my struggle who's out here, a comma and he passed another. My name so.

E

B

Daniel yeah: that's how I missed it for America ended today.

E

So this is the observe my predicted over epsilon hi, and we thought that the expected value of this was above equal to n the number beta. So we were looking for a fit to the to the data that gave this to.

B

Being something in the order of I.

B

E

And then on our five D plot versus five Oh.

A

There's another one here: I think.

E

It's me so remember: we had by d5m.

E

People, so this was fine, the scar and fortify the star was approximately constant and the number of data which was like a hundred and sixty okay. So if I put kind fashion equal to one, then that's what I'm looking for, if I meant to a one with Tennessee these.

B

E

I put kite fact.

E

Equal to tan Talib, dollarization, okay, yeah I'm, actually going to be looking for a target misfit, that's equal to eleven hundred and sixty okay.

B

I'm good to say, I, say no yeah.

E

D

Let's, let's just.

B

E

Let's just put that and then we're going to run it excellent.

B

Anti-Drug runs Batman.

E

And here it says now: I achieved a target misfit of eleven hundred and sixty days and I.

B

Thought how can isn't that about te water, perfecto, says well and.

E

This is eleven hundred, so this was our argue fault. This was equal to and 116. We've now set our target to be up here, so we're stopping here and if we look at the data and at the model.

E

E

Here's our final bottle it's at a hundred and forty two for the for the blue. It's not a bad shape.

B

God, don't do that the shape that it goes on yeah, sorry time, mother, Kali's and ara. So.

E

This is a simple model: just a block in a half space have.

B

Spaced them out, we had a lock of the door for anyone. We.

E

Certainly haven't fit the data quite as well as we'd like they.

B

Died off it did nothing what's happening so.

E

We've got observed predicted and there's quite a few data in here that we're not even on the scale of plus or minus Li Wei.

B

E

Right so it doesn't even fit between minus 3.

B

And plus 3 so minus beneath custody, my own finger tailor. She got me kidnapped, got damiana so.

E

In situations like this, okay, we feel well. Maybe we should fit the data better with.

B

Unoffending double and if.

E

You want to fit the data better know the dates as soon as I know. Then we could reduce the kind and backwater of Yahweh away and now run this again.

E

Hey so now it's achieved that.

E

So now it's it's come all the way down here. To get this target. Misted, mystical.

B

It's he will drive a long time. Okay,.

E

So we expect that it's going to to do better. So let's look at the model and at the last iteration.

B

E

Look at this about a year, so the picture doesn't picture kind of looks the same.

B

E

At this value, this is 50 to nothing Nick. The previous one was a hundred and forty.

B

He gonna have to gather that Nick so.

E

This sum this sometimes happens that you, if you under fit the data you get poor amplitude levels for your resistivity, assists.

B

Even if I live it I miss fit for their. You know and.

E

Also be a lemma yeah and also in more complicated problems. You don't quite get the rights.

B

From that in the novel ratio between ahead neighborhood again, if I get it, but we got some at me.

E

Something that you notice with this I think maybe Devon had pointed it out earlier.

E

So look at this curve it's coming down and it's.

B

A neat way today, but it's.

E

Coming straight down, so it looks like you know, maybe maybe that's easy to to fit, in which case maybe there's information in the secret in the data that I haven't pulled out. Yet there's.

B

Nothing diabetes, auntie, daddy, hey daddy! Forget it t that you touch it do ya uh-huh! So.

E

Let us let's try to push it some more. So, let's go down, let's spit it down to a KY fact of point. One got.

B

A little kind back, we drove down to down to mess it up over out that machine. So.

E

E

This so now I'm going to do a kind fact of 0.1.

E

E

So now we're asking it to fit the data really well, but again we have noise on the data.

B

They don't fit the exact event event. Do you know my leg? My eyes got bad as well, so.

E

You know there's a question of whether we can even fit the data.

E

So here I I'd, let it go ten iterations, so.

E

Look what's happened so.

B

E

It's kind of come down here and now it's a plateauing.

B

Dnd one, so this is reading.

E

B

E

This is an indication that you know, I've probably gone about as far as as I could go, and you know maybe um maybe it's worthwhile looking at say a model and ears, so maybe I could look at if that's Tam 9. Maybe we could look at that long.

B

As on a clock generation, a claw machine at modernities, yeah.

E

And you see what's happened here: we've actually got a more compact body and we that's that's- had a resistivity, it's even lower than that within the true models. So.

B

The reference already the true morning, taken up over an angle in.

E

This case, the.

B

E

Model was 50, so it was a block in here good.

B

Morning my title now say: welcome laughs, any kind of the month and how they have done it. So.

E

E

The lessons here, but go back to 6.

E

E

Oh I see I sorted by the color scale.

E

Okay, so the the lessons here are a couple of things. One is. The Chi fact allows you to solve it's an easy way to find models that are a little bit better fit or a little bit worse, fit in your inverse.

B

Problem to us, it must brother that monkey I bet that setup. We didn't pull down on it.

B

E

Why is that important by.

B

The base at all, it's.

E

Because of this, when we solve that inverse problem, we're setting up our observed minds predicted divided by epsilon hi.

B

These are uncertainties and I would get into that and.

E

We never know what they are.

B

E

So we always have to guess: oh, come.

B

E

Usually, what we do is we guess that these uncertainties and then we follow that misfit curve, and then we yep we can choose a model by adjusting the Chi fact values.

B

In fact, an important rule- nimi Rama ie yo man, I'm starting a TV camera ma ma ma who cannot exceed a mystic named Wayne Airport on the China. So.

E

The contact is just really an easy way of kind of adjusting all of the uncertainties that you're applying in the data I think it's utter you have such an integral. He.

B

Didn't let, though,.

E

That well so it's a very useful kind of parameter to have to just adjust how well you really want to.

B

Try to sell but I'll keep it inside, there's no nodes in it. That one was a very tight back door.

E

Okay, so those are I think we've now explored most of the parameters in the inversion. But what I'd, like you to do is to just experiment with the simple models that we've got with the block model: a lot.

B

More inevitability and I would lose Lancelot a lot. I don't mean you never lose three out of your care as well as we're going to do some types of time, because.

E

The inversion runs quickly, and so you can change your parameter and see the outcome so.

B

I've got five me that intramural veloute yeah they try little delay in any category, looks about it. So.

E

That will give you a lot of intuition about how the inversion works and then, when you go to the field, data you'll have a good understanding about which parameters you might want to change.

B

And the military version it's it's have to become about it, be that what I got out of looking version for Department that all Tom never little kid no seed as well. But this.

E

Has definite is going to talk about tomorrow when we do you know the real practical inversion of the field of data sets. There's nothing really. The changes between what we're doing here and steal datasets depends on.

B

Time ago, man hogging the in question about the moment. The game didn't know, particles anything wild about it, except.

E

That the problem gets bigger and that means did you take more memory in your computer and each iteration takes longer, whereas here each iteration was taking only a few seconds, it might be that each iteration of your field data set takes 20 minutes or half an hour.

B

Iteration, young mother, not my worry, there is a different matter. No. That was that if I got really got that up, look at me and hold on elevated are you got me? I got the iteration unit of Zama yeah yeah.

E

So we're working to make the programs more efficient, but even now we can invert word your data and for those people or the DRD that have got a good pc with computing power you'd be able to invert those families.

B

Or print a Thomas and I think that's it.

E

So anyway, work with that to block or with that block to D block model and experiment with all of the parameters.

B

E

And before you before, you click the Run button run.

B

Run for that global or naked yeah.

E

Just anticipate what you're going to get I want. One more comment:.

E

After you, after you've run an inversion, the best thing to do is to unclick this Run button. The.

B

Raul be white, knuckle, give demonstrate I'm, Rumble, unclick, I'm, clicking and.

E

Then you can make whatever changes. You know I Ahmed in here when you're satisfied with that they're, not like you saying they.

B

Don't like the rock around any man.

B

E

So we only have tomorrow.

E

Devon is putting together some movies, we'll talk a little bit about the data inversion and that divots.

B

And no edits at me on hand. We do a segment on max.

E

You think we still try to capture it's worthwhile capturing you talking through the iris system.

E

So we might do a review of the iris system and also try to capture that light on on a video. We're trying to make sure that when we leave here, you've got the resources for all of the things that you need.

B

You've got about vowels as well as well yeah when I come on yeah, the one that you need it. It's a game. I serve I got ten o'clock. We live the goombas I'm somewhere. You see me play uncle. He is they made lot of those happens in a corner. So.

E

In that sense, you might want to ask us questions tomorrow to make sure that everything is clear in your mind, because next week, you'll be going out on.

B

The road Venus you gave me this I'm, not that was up for.

E

Me well no keV, it's here, but you're you're supposed to be well Kevin's. Here is the expert right. So if you have a question- and you ask it, but really you should be planning to like I'm, independent I could go ahead and do this you could get into problems than Kevin's there to help you and then later in the week, there'll be another person from Boulder off. Also here.

B

Chairman Tang Amazo immunity, okay, but I'm alone, as well. No more in this muddy mess with anybody animal but.

E

I think, from the perspective of the DRD and the main bar group in general, you want to be as independent as you can and as a group have enough knowledge that if you're out in the field that you can solve all the problems that you need to solve. And then, when you come back that you can know.

B

What people are you teaching them a process? I.

E

Don't Kevin you never and Max anything else. That's max.

B

Photos from with that, oh yeah.

E

I'll put them.