GeoSci.xyz SimPEG, 4 Apr 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SimPEG Meeting April 5

Description

Seogi updates us on grounded sources in the TDEM code.

A

Yeah, so now that there's a couple things I wanted to update you guys on just quickly before you get going yeah.

B

So this is a cup.

A

Does anybody is anybody else, no.

B

um It was going to do is just I headphones.

A

Feather okay, um so one of the things that is coming and I just merge. This Rowan worked on the examples I'm just going to share much thanks. You guys it's.

A

So we now have an example gallery. So if you go to it's all, I've done and get dogs once we get this into master um that will be gonna, be live on like.

B

A normal dogs make sure sure.

A

Cool, um so if you go to examples now we have a beautiful gallery of synthetic examples. So perhaps now I'm deploying you so thank rowing for putting the time in to make that happen.

A

So now you can download them all as like a zip or notebooks, and when you look at each individual one fits again folks: options which is really cool.

C

um So it looks like we generated, like Sophie update, for instance, like that example, multiply the metal regenerated this one yeah.

A

So what you have to do, if you go into simpe, is the examples, have moved so they're now here.

A

So the examples are sitting here and so the broken down into a basic mesh. That's all of these guys, so you just create an example wherever it's appropriate and that's all you have to do so, there's no more running of that and knit file or anything like that. He just create an example.

C

A

John and soggy I know you both have things to share who wants to wants to go? First.

A

Don't you I, do you want to take the lead here? First, I.

D

Sorry guys, my connection I, don't know why, but it's really lagging I'm, not gonna, look gonna, say anything I'm, just gonna try to hang on to you. Okay,.

A

No problem: well, then, I mean you're out.

C

C

B

You, like the the chordates you wanted to share your screen and I can display yeah. It's all just share this.

B

C

uh Justin yeah.

E

C

Yes, I got some updates foreground resource for time domain en package so like previously, we didn't have that initial condition is implemented for Grandy's horse. So now.

B

C

Initial condition is, was happy depend upon the conductivity model, so you need to take the derivative and then you need to put that into like your code. So.

B

C

Did its as murdered yet that it's in the a.m line source branch, but basically you can suffer instance if you want to use that line stores. So.

D

Sorry wicked, oh we're not seeing a saiga screen.

A

F

Need to click on his feet.

F

It doesn't pop up automatically because because he isn't sharing his mic, so I don't know if that gets recorded to the video yeah.

A

It is okay, can you see I know? Oh.

D

Yes, thank you guys.

B

C

Yes, so there's a like a source is equal to en this. This fine is basically like that's what you need to do so blind current, and then you pass the current path. So here, while you can put pretty general path like so, if you have a pretty well defined.

B

C

While I'm path that can handle like those patents, so it's integrate over to the edges. So this is very easy yeah, it's only a relation to for a change you to rewrite that code and it's only work. Four.

B

Pence Ramesh put.

C

Up to english, okay.

C

Okay, then right so I ran in version. So it's pretty small example. I think the mesh size about 150,000 and then I got tenth on channels. I guess and the data to can recover. So this is a true model.

A

Hey quick question: why only at work for audrey? Is there a certain assumptions? Yeah.

C

Well, like a wedding there's, the only things like if I can find almost like a crisp on the cell, so where they're wider path goes yeah, because I generate the sort of interpolation basis and then integrate over through that. So I need some functionality to find okay ever good to pass on that path, and I need to know which cell line passing through. Oh.

A

It's a chance that, like if you were changing cell, so you just like straight up: miss Brooke.

C

Is like I assume like what to write it, but I assume the tensor match so I yeah, but it like if you got some agility cuz. I think it's not like a major change to write that but yeah. So as.

E

Long as we got on the same.

D

A

D

C

Month we might get to go back at light, yeah just.

E

Putting it at anything like.

C

That that's what I'm put just off your national run.

B

C

E

But if you had an octree match for the course center of it was the same as like the coral tensor mesh. Do you think I'd like to find it on at intermission and route it on me, optra mission, all the edges and stuff we're insane or patiently didn't angel yeah.

C

But skill, then you need some sort of gap in index. You still me some a pin code, we're like because it's all about like indesit, make the values and numbers on the edges, see if you can find the crust on the edges in the alchemist you're good to go, but but I wouldn't I wouldn't do it. I was just generate in the up like 30 in X I we would be writing code, but yeah I think it's not a huge deal, so it can recover conductors and resistor. So here I am.

B

C

Http to generate that poor law and data- and then it's like compared with that that BBC co in and that's just pretty well and I- can recover the model as well. Sorry's can recover both conductors and your sisters very nicely.

C

Yes, I didn't do any fancy things I. I think I just put the yeah.

B

Default inversion.

C

Parameters, so it's about like.

B

C

I guess cool, so that's a basically looking at so meaning like you ever Great God, but on the source data it can you hide that works for both the general way plan on the stem only one now so. Oh it's.

B

Pretty done enough.

C

I think that's it again. I got the.

B

C

Example with the DC example, you can get better better research, which is sort of expected, and this is one read like a worksheet. Why? Yes, I think that wire pad I think I send an image before I, yeah I, don't know, but.

B

My wire path is.

C

Here, like this, okay.

E

There's no menu kind, upgrading or a data yeah.

C

Yeah yeah, that's are my favorite simple. That's all example so points them, but for these t you don't have that resolution in this case, so you need to sign the best way. Dima put things in them. Yeah.

C

That's my that sort of update in the time domain, yeah.

D

C

Things I'm. What I'm working on is like a processing problem, so, for instance, like say I'm setting up like can it's a time that behind became curve, so you had.

B

A whole bunch of.

C

Time you came curve in that your warranty, for instance, yet and I want to fit that itch itch stations, but I want to kind of put some sort of constraint on the states. So sorry, if you want to fit.

A

It in terms of it was a symbol. It's a single, the Cape. It's.

C

Fitting exponential- yes, basically I'm, not betting, exponential yeah, but the life got that slope. I can I want to put some constraint in space okay, so we can set.

A

C

An end result each station separately, but you can set up a inverse problem together so and you can put some regular this each. So it's like a kind of server different use of what like.

B

D

How you can use this.

C

Impact inversion last so basically like this is the wave form that I'm putting in for p.m. and so I. It's basically the same, but are you input fully warned data? So I think it's a said, mom really good example, and we can.

B

Run the exact inversion.

C

And fit all the curve yeah.

A

C

A

You're sitting all of data in okay, so.

C

And then, at the end of day you can get power mad and the time to me like so can okay, so like a palm app as well. But you have you have a multiple towel here, it's a little bit different, but anyway.

B

C

End reserve can get the pal Matt, but you can get softer proxy of uncertainties because, like you're using smooth basis function, yep it that it's a simple processing code that you can simply use so like, for instance, the how Matt looks like that and then I was using like I. Just I want to pick just a positive day, yeah sure, and then you can simply be couple. Ip lies like as a possible yeah, so you can recognize where your IP signals are yes, but it's a simple curve fitting road but I put.

B

C

Meta sponsoring so, for instance, like how the code looks like it's sad. So that's looks like that.

C

Yeah so I have a like a survey. It's exponential fit survey. It's very simple yeah you can said: I need some like a way to set uncertainty for my case. I've had to put it in, but generally you don't need that. So.

B

Basically, like you.

C

Just need to be like a ball and then appears on problem. You have a different kernel function that you can use, and basically here I'm generating the sensitivity matrix. So that's just months, then I don't need to update that. So, if it's in your interest all right, this is pretty to its same, like our structure same so now: jay bag and JT bag yeah, it's pretty similar, but flights now I've got G matrix, I think probably thumbs. Mecos might be pretty similar to this.

C

Yes and then like my line, regularization, for instance, for airborne case of each line. So.

E

I can put a smoothing.

C

Constant in line yeah so.

A

C

Make it smooth in line so my flight crews, inversion primer will be something. This is basically all.

B

C

But much simpler problem so yeah, but you can rewrite that, like just as part then you're 30 50, pound yeah, then there could be multiple application. So this is a one case, but I could do it for DC, like a DC, IP or instance,.

C

This is a peek sample.

C

So this is my the kicker for IP, so I'm calling over.

B

C

Courage that I have yeah- this is a viable bike. Word all right and look at the pseudo section.

C

Basically might yet, but I Elvis showed me the research. So if you look at the so this is apparent resistivity to a second, but that's the data. Ok and then that's IPC row section chart of the way to session yeah, but if I give multiple time channels, so you can fit that and recover way. How so.

A

Sir, the apparent treasure believability ears, showing is a snapshot in a single time. No.

C

Yes, ah no, it's integrated values to do what they say: Leah, integrate a certain time, channel yeah and then walk today above like that sweet window, but a play. Let's give a multiple time travel, so you can actually fit that, but basically same idea, you can use exponential function or whatever, like a simple function, to fit that curve, yeah and then recover.

C

Some products, I haven't fitted in east movement option, but you can have to grab some point and put it in as well yeah, but end of the day what you can get out of it is like, so you can get out of like a multiple progress from that curve. So this is a see value. This is time constant. So time process is interesting. It's like you got stomped ran, but the again like this is sort of what the thought you can get its across. Your mom is like this is thao.

C

The x-axis is how and y axes a parent chargeability. You can see the trip Tran, but you can drape over the paramedics, tivity, meaning like a typical region where you have a small power and small target ability. Yes, but be like a high resistivity, this region as and the greater charge of the way so.

B

You can sort of like.

C

Some source classification, but it's just the only data space but yeah like it's a simple data, fitting all group. Did you came yes I? It's still. We like the thing that I was thinking because we don't find have any data structures yeah. So that's probably what like we yeah at least just a table like structure. If you can have some simple data structure that we can consistently used again, but probably.

B

Because I um yeah.

C

I'm doing is just I'm using just a dictionary at this point, but like some like that great for DZ.

C

That's sad for I am and.

A

And weird, where is this.

C

Is the Sun? Oh, it's.

B

C

My local branch at this point yeah but I, want to is.

A

C

Happens in Dakota, it's called sink IBM IP flat. So this is where I'm putting most of my face, you search yeah, they're, more like they're. The main thing is the hormone encode and yeah and put ya Pete eggs and form aholic, and but like there's, some tility coach that, like simple processing code yeah, go.

B

C

The color sometimes yeah, hopefully we can version one yeah.

A

C

Just still actively working.

A

On it, so I don't like you.

C

And father: okay, no.

A

Absolutely I'm just curious cuz. I wanted to sort of see how you some things through the problem, in particular. Getting some ideas harm the data structuring about you to look like yeah.

C

But I'm really really like. Is that not sure how you guys, using this wires and and property time? So I got like this: oh, no, this property yet and how I'm like this is like way, I'm putting things together. It's uh there's my wires. Okay, this is my wires yeah.

A

C

Yeah like this is a simple case, like all of your variables, are, have a same dimensions: yeah but like depending upon how you formulate it like it could be different. So it's kind of cool at the end of it and use the same mapping to output the how we are using like because I you have a big long vector. So you need some matrix to extract all your all your variables yeah, somehow yeah.

A

So that actually.

E

C

A

B

C

Pretty cool any clearer in a.

A

Combo regular edition no.

C

I on yet baddest duh, that's a my nexus.

A

S step: do you have a small example that I could help? You is right? Yes, that would be yes.

C

Sir yeah so I'm sort of rebuilding couple things in in DC and like a yen code yeah, so I could have a small town. Oh cuz,.

A

B

Awesome I, curiously baby wrinkly, Donna.

A

hmm That's really cool Tony yeah.

C

So that's how I'm, imagining like that kind of processing stuff, because I at the end of the day, I think a lot of processing problem can be set as a campers problem. Rather than not a new problem. Yeah meaning like if you got.

B

Certain data structures.

C

Yeah then, like we can actually use impact framework, get to process it again, yeah, and so is it a little bit differently yeah and this them now. So that's what I'm you thinking, what.

A

Are you thoughts on developing the processing as a separate package work? It should be in soon peg, I.

C

Think I would just go like, for instance, like I would just focus on the yam now yeah and then just do it. It yeah, okay, so yeah, I'm, Tom, processing, yeah I. Think that's fine, because I get very like align with certain problem. Yeah having a really separate like a package is kind of I. Don't know I'm not sure yet.

F

They're, honest a great.

C

Cuz like and then like crossing bankex force you to have a data structure yeah, then that will be helpful for in both weights, like motion, processing and also in a synthetic problem. Yeah I. Thank you serving yes, ya know.

A

I a degree I think I think originally we were thinking of having it separate, but it it's so tied to the problem. Yeah.

C

I was betting on, even though.

A

C

Four empty processing I would attach to any CM Punk yeah. They may cesium Bob Costas yeah. First, something like that. Yeah that might be more. You could probably have a more general structure yeah, but he at the outside I think like p.m. dr. house in water, pyromania, ganbatte Austin. It that's kind of make sense to me. Yeah well,.

A

I think like in a lot of ways developing the same way we did with disc retirees, basically is. If you develop everything in one place and then once it's for sure enough of you realize.

B

It is a self-contained.

A

Piece, medical, but for now, like it's part of its part of the IAM, so yeah I, think that makes sense. Yeah.

F

I read yet we tend to target a different okay.

B

What are you x3 to.

F

Actually develop it outside of sin pack: okay, because I mean it be, I mean it doesn't really matter, but it's developed, but it should not necessarily live within this impact, because impact is fairly focused on the modeling and and we do have a data structure within sim pack, but it just. It really needs to be reworked, because it's it's very geared towards the inversion problem, yeah or the or the more modeling problem so and.

F

Yeah so happy and.

C

F

C

Kind of group is like that I'm, okay, with where he believes, but I think that we meet some cuz I didn't surf. Thinking about on using again cuz I would like I would prefer to do like okay, like if I'm garyx a mt processing, an inversion. If I could do n, SZ m dot, processing, dots, blah blah blah me.

A

C

That would like resonate with me if I, maybe I'm, not sure having quite be in that situation. So what I'm doing like now, I'm using DC and IP in these days so well, my workflow I had to put my data into certain framework. It's still not I had to generate the data data frame by myself, which creates mean basically dissipate ball like locations in the data type ion channels, and that's that's just one, my work Lloyd you can your processing and then like it need to make it as an invertible.

C

It then get invert I think that's just a whole workflow and if that can be in a one like at one or like one shock, work on is I. Think that might be good to me. I look like I'm not sure like where it was a beeping, but they need to be accessible.

B

C

Interest single into single frame: ra, that's what I I was anything ya.

F

Know I you on that soggy, but I would say that we should have were kind of own or processing and I mean data in general should almost be separate from the inversion and modeling site because you're dealing with, even though it's you can reuse some of the techniques and the in the infrastructure that we have in its impact. It's very different approach and yeah, something that you should I mean you should bear in mind the fact that you're going to model it or you're gonna invert it later.

F

But it's also an independent stat that needs to be dealt with. I would never take field data and process it and then in verse it kind of in a single kind of pass. You would do the processing and you would kind of make sure that the processing of data is up just after standards or up to snuff, and then you would proceed to the inversion, not saying that you should shouldn't go back and and reprocess, but it kind of it. These are I, don't know two animals that live in the same sooo yeah.

A

And one of the things that I think is interesting, like from your first point, katie is no.

B

A

So far as focus a lot on the modeling, but in some ways I mean we, those.

B

A

From the inverse problem is the primary goal, um so it's it's a question, I guess of scope and then also of strategy, because in some ways like I do like the way we developed district eyes is that it? You know it. We tied everything together and then once sure enough, they pull it up, um but I.

A

Don't think, there's married to both approaches and I'm sort of wondering it's the strategy to take care to figure this out, so we're certain you're not going to resolve it today, but it is to sort of lay out like what are the types of problems we want to tackle in processing and like what song he's already gone. Just like have a list of you know things that have been done and we can start to just get a feel for what this looks like and then maybe make next step from there. Yeah.

F

I also think it might be easier to get people involved in processing. That's fair without kind of getting into the overhead of sin. Peck of the modeling I mean it. People probably will want to do both, but it might be kind of it's less daunting to get into one thing when it's kind of its separate, but it's kind of like that impact sphere. It's all connected, but it's still like discretized like it yeah it's but I, think just it. It is such a.

F

Unrelated related topic that I I personally think it would be better to kind of keep it separate. I've been thinking even with the energy and thetac lat, whether it should be because then you're also not limiting yourself to only simple, but it it it I mean it's obviously optimized force impact, but it but yeah. It's just kind of question how and then I guess and with empty pie, for example, it's kind of how are we going to interface through something that already is there? That's.

A

F

I mean III think it it kind of connects to a bigger picture, but I think we should, but at the end of the day, it doesn't really matter where we develop and how we do it initially, but it just need to be mindful of where we're going with it.

A

That's very different.

D

What you got there soggy? What is this.

C

Okay, oh oh sorry, ah oh this is to get hurt. Judy.

D

C

Got some comida, so what I was doing I did that to the inversion first, so I got the four lines of dipole-dipole theta and then do the 2d inversions and then like doing some decent. You see that I could fit the cover tomorrow, as I play with some weighted stuff like yeah. I also did the premium.

B

C

Is my I got a pretty poor pong team so but I recovered like some resistor and inductor yeah, so yeah just like it was actually my first kind of like a little serious three inversion using synthetic, so yeah. It's like works, pretty well press up to know, but actually works.

C

But what might be good to develop I like side, so I get it just by myself, but II kind of a good kind of survey class where you can extract the TUI lines and kind of from the trig inversions and like or we can run free conversion or you can easily visualize that type of thing set. So I think that's not hard but like just need to put it together, but we got all the pieces like that got like we can do a free over to the topography of the electrode and those.

B

Can functionality be 45.

C

Because that it just speaks your weight like I put it together as a good kind of work, phone yeah or at least the functions that we can run to classes so.

D

Yeah I saghira you sent me a note book is that is that that one actually I can't find that you, mine, resend it to me. Oh yeah,.

C

Yeah I'll sent you ah right. Yeah I think that yeah I was using some utility stuff, ha so yeah right. You.

B

C

That's true: ha that's cool, so I, I kind, okay! So here's my question and do we need to put like those function already in the survey class? Is that working what we need to do or what's uh sorry which functionality? For instance, okay, got three depot three days: yeah yep important song, yep and then I flip it into a syntax, rivet class, but I got all the lining. This is like for all of the my electric vegan I got mine in this, so I can easily find which line?

C

It's is it if I can generate this 2d survey in the survey like a 3d circuit class? Yes, if you want, if you can't have some functionality, that I taste that he surveyed and outputs that you Ilan it I think that's simple, probably I mean life visualize it in.

A

Some ways and potentially like to see it a level lower, is that we can basically bring in the data and then be able to sew on the data clocks, be able to generate a 3d survey or crowd of line and generate a TV service to actually do it from the from the data class. And then surveys can be generated on top of that yeah. Okay,.

C

So currently, all are like a visualization, for instance this guy or yes, it's like mate 8e furnaces. This visualization. We take the survey class to generate the pseudo section yeah, which is okay, but that's a little bit different. What that you mentioned busy, so yeah yeah wasn't really sure what.

A

C

That what was what is this way.

A

Yeah I think like rethinking the survey and gave us needs to happen shaving, um because one of the things I encountered with I got email from league who was a student with young and she's working on generating time to make I wants to form all with both on activity, and so the things she was doing, which like makes sense as a first house, is that you know she was generating all of the source locations and that all of the receiver locations and divining once worst, which is, of course not how we do it.

A

We generate you know each source is only a single occasion, but we had no error handling to catch them, and so she generated the director of sorts locations and a vector of receiver rotations and so the way yeah. We just need like more structure around helping people set up series. I guess my life on it is.

B

For generating that airports or video actually just had to generate.

A

A source single receiver for every location, rather than doing the list, but that's not really it's not really transparent from the outside yeah.

C

F

Mean a lesson I can't imagine it no.

A

I thought up thats easy, but it is like it's not it's not hard once you know what to do, but if you sort of it's just your first encounter asim pega, it's not right immediately obvious! That's true, sir give me you were gonna, see someone yeah.

F

I was just going to come with the data that I totally agree with this I guess: I checked out kind of Brie formatting or revamping the survey receiver data interface, but we also need to think about the implementation in terms of what is the data and it should be flexible, but I mean we need to be careful, how we, through the inward problem and how we need things to be set up, and then we need to have something else, that people can kind of more easily take a data file or a data survey or like their work and make something happen and kind of on a higher level without having message related to dive into that, we call it receive earth with an ass, meaning that it can have multiple locations versus a source is a singular and shittim only have one location.

F

So just I mean these are details that if you just have data- and you want to work with it- are just annoying to have to go through, you will probably want to do it after you test out your data. But it's just it's a it's a it's an inconvenient hurdle to have to jump right from the gecko. That's.

A

F

And I mean same with the data structure. I mean the data structure is because it's referenced as a source like you reference it with a sorcerer, stupid pair yeah and I mean for empty. That makes no sense. That is just not how you would think of your data.

F

You would think of it. I mean you think of it as a station, so we contain all the so it's kind of location-based, more than.

F

Yeah, it's kind of just having having flexibility and that's it yeah and I. Think I mean we can definitely do better on the receiver source implementation. But I think the data needs to be our data class. Isn't it to be kind of a good gateway for people to get into it at a higher level, yeah and kind of build. The reason functions that are needed to get they die in, like it's such a nice place for people to start contributing very.

A

F

A very difficult place for us to start to implement something because we're not used to be working with all the data.

A

Leaving the synthetic world love.

F

A

F

Yeah, those are my larger might depends on that, but I think, like we've talked about like the location and, if somehow being able to protect the locations as well, because I mean in the current form and we're storing the the I guess, the production projection matrix it for every single source, receiver, combo that we have and they're like repeated I, mean in the empty class I mean if I do even do your synthetic I mean I'm repeating eight times for each shorts so and I mean it's also question whether we want to actually just think about I mean constructing some sort of Val, a sparse matrix that is just stored in the background.

F

That would make the projection a lot faster instead of forming it with a for loop, because I mean it's not a it's, not the bottleneck in the code in the empty cup, for example, but I mean it still takes me. I mean 30 seconds to project I mean a really big survey seconds. That I feel that it's.

A

Not necessary my.

F

Gut feeling is that we should be able to do it in a lot less time. Well,.

C

I sometimes I do cheat, so you can like a can, make a same receiver class and then pass it through all this all those like I sources yet, but at the end of the day you can assign the infinite amp it up in to your certainty, then, in that way, like you're, using like your it's in that inversion, but like impact and your inversions, oh so, like I, think about like if you're a certainty is infinite.

A

C

Actually, zero playing Kurt, okay.

A

Sorry, what do you achieve by doing this, but.

C

But you only get it one can decrease or like, even if you have a hundred sources and then you're not like a roofing /, predictable you're, not rated different receiver class. So you can find all the unique can find unique with civil occasions, yeah generate one receiver class and then pass it over to all the sources. Okay right at the end of the day, just need to know which is not included in that sources.

C

So you need to know that emphasis, but can assign that internet to certainty so well, I think that's actually how the ubc put works. So they like they find unique serial locations and am generating 1p Beatrice yeah for projection or evaluation.

C

Then then they generate that, like the weighting matrix and then they put you don't like that summit and Valerie work like zero on the double so in that way like just need to generate one piece or but yeah, we can still do it like in, but yeah I'm, not sure which I think we admitted that 1.9 discussions would go about this, but the I think we choose different way, but still we could do that exactly same thing: ebony, because structure yeah it's up for the pack, ok, but like that, but there were some classes for cool, so they can.

C

Even if you generate one superclass, you can actually pass it over to multiple transmitters right so yeah that doesn't like a walk you to do different things yeah. So.

A

You don't need to generate the.

C

Servant class every right, yeah.

F

Yeah I mean it's, it's just worth exploring. I think that's something that is very worth exploring and kind of consider more seriously. Yeah.

A

Absolutely so, we've only got a couple more minutes here and wanted to just quickly go through there's a few full requests on the go and just like quickly chat about strategy for getting things interesting share. My screen.

A

Ok, so the docs are now in so that should fix deployment issues that we were running into there. um So there's a few different. The next one I think to go in is going to be the objective functions, um so I'd like to get that in hopefully like by the end of the week, because then dawn that'll. Allow like me to do.

A

A sane will request her you on the mag amplitude one because it is they are sort of intermingled, and so we can jump in and like start getting that ready to go so I'm gonna see if Rowan has a chance to take a look at this dawn, were there any concerns you wanted to bring up or like thoughts on implementation that you think should be um yeah. Give it a second pass before we bring this in.

D

Well, there's the issue of having two levels of over regularization objects right, so we have like either a combo or just a sorry. I forgot the names, there's combo and was the other one.

A

Reese objective function, yeah.

D

Yeah so, and and that's where I added a bunch of stuff on my on my PR to be able to deal with those, it depends. If you want to do it now or we just do it when we merge, like amplitude.

A

Okay, I'm wondering actually, if the thing to do is to generate to switch the base of your pull request to the RAF objective functions um and then that way, like I, can sort of. If you can point out an example of where you're doing that, that's something that we can try and fix before we bring it in to death. Ok,.

D

So moving yeah, okay, I see so I'm moving to a PR to inside your ear branch, yeah yeah.

A

So it should be like justice.

A

All you have to do is edit and then just change the base to ref objective function and then, if you kickin it just in the files changed, if you can just leave a couple comments of like hey Lindsay, this is where the objective function. Class inheritance is messing with me. um That would be very helpful.

D

Yeah sounds good. Second, I can try to do that.

A

And then I no posts, soggy and Kootenai have clerk west onto e.m dev to you guys want. How quickly do you want to try and push on those as their stuff? You still want to do.

F

C

Yeah I'll do it.

E

C

Weekend I like I mind that many thanks to okay, cool I'll go to your comments.

C

D

Granny, a natural source I for some reason, I I, had to look at the 1d, the 1d, inversion, yeah and doesn't look like the regularization is doing anything should parameters and pretty much always a sink on virgin. So always the same result. What's what's happening with that, I I.

F

Honestly, I don't know it's one of those things that I implemented kind of when I wrote the code and I haven't really looked at very much so I'd have to have a decent look at the 1t problem. I. I remember when I implemented it, it was had something to do with the active implementation, but I mean that was before we had like the active in the objective so I. I I'll admit that I has to look at the 1t example a little bit better, but that is part of the work.

F

I did I've re-implemented the tweety forward problem, so that is what I'm pulling on or trying to push down to en dev, among other things. Ok, so I'm I'm really trying to kind of just clean up some of my work space to just organized, because I had I had like seven different branches on my computer they're that I've had to kind of merge into one another to get things to work. So I spent some time yesterday, killing that but I they're always falling on some doc, don't understand!

F

Okay, so I made that I I'm gonna cut the en e m merge branch, whether I'm just I'm, taking the e/m brent and merging it back into natural source en before. I then merge that into the e/m branch. But that's one other circular.

A

Okay, oh, if you ping me when you like, want somebody to jump in I'd, be happy to and like because if you can resolve or I guess, there's the Travis fail enough because of the dogs and then there's some conflicts. Yeah.

F

So I think I've I fixed the conflict, and I am now just read running some of them trying to rerun the tests. Okay, um yeah so hopefully yeah once it passes, but I'll open you if it if it fails on the same error, it's something about not finding a package. Okay,.

A

Hydro, hopefully fix that on the most recent exam are the most recent docs deployed, but it quite hostile not so yeah.

F

Yeah but yeah so in terms of natural source, the state of the code I mean it's I've, been kind of I have an issue going kind of a general issue of development of the natural source code. So a lot of or some of the things on this pull request are things that I've chipped off or picked off that list perfect.

F

Don't what me it's a better mom yep yeah.

A

I gotta run, um but you guys feel free to keep chatting if you want to.

D

Thanks NZ, that will see you later see.

A

You later ticker, hey.

D

So I you sent me into a book like a bite.