DevoWorm Lab Meetings, 10 Aug 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: DevoWorm (2020, Meeting 28): GSoC updates, embryo networks, Bacillaria cognition, Chaos measurements

Description

Attendees: Bradly Alicea, Mayukh Deb, Ujjwal Singh, Susan Crawford-Young, Jesse Parent, and Richard Gordon.

A

Hi, how are you good.

B

B

C

How are things going for.

B

A

D

D

D

A

It doesn't sound.

A

Good, hopefully uh yeah we can get some rest.

D

Today, like much better than yesterday yesterday,.

A

Yeah, that's good.

C

A

E

C

Jesse and have a.

E

Okay, um I think you can hear.

F

E

C

Can you hear me.

E

Yeah, you can see me, you can can't hear me, but if you can the camera and the sound seem to work off the same circuit in this computer.

A

E

I may need a new computer. Oh.

C

Good, maybe I don't know but yeah well I mean we can hear you that's good.

C

Welcome uh so welcome to the meeting everyone uh again hi jesse um we'll have uh some updates from usual on my oak in a minute I just wanted to go over, maybe some other things that are going on.

C

We just submitted a paper for a vassal area behavior, so we submitted the abstract for that and I'll talk about that. A little bit later for another book chapter, that's being written the proposal or kind of more of a it's not line. I guess uh they don't.

C

You know it's just based on what we had. We talked about previously, so I'll go over that in a bit um also, if anyone has any news um on their side now's the time anything going on new, we might be of interest to the group.

G

um I mean- I don't know if you probably already know just to say this for the group um looking at the two papers that I mentioned in the world slack channel, like I'm, I'm interested in looking at those papers again- um and um I mean I'll- be um I'll, be starting at summer school later today, too, um which will be interesting on brain's minds and machines.

G

um I don't think that very much, but I'll probably talk about it next week. After uh and lastly, uh there are a couple papers in.

G

Like around that's kind of the match, point of neuroscience and cognitive science, but um like bradley mentioned one.

G

Like biology at scales like that, that paper- oh yeah, definitely in that paper in particular- so I might go over that in the future, but not this week.

C

Yeah, well, that's good, yeah, uh susan, of course susan uh and I we were talking about the uh lotto embryos last week, so any other additional thoughts on that or.

E

um About the accidental embryos- I I haven't done much this last week because I was kind of given a reprieve on my thesis writing for the week, so it was sort of a off week.

A

E

Trying to straighten out my office partly um I need to- I would like it functioning properly for september, and that includes having um having my 3d printer working so that I can make a ball microscope wow anyway, so uh yeah it was sort of an off week.

E

I haven't done much um as to the chaos papers uh and I don't know whether I'm going to use it in my thesis or not, but it is a method of measuring things that seem unmeasurable um and so that that's why I I thought you'd be interested because it is a measure of behavior and of um how fractal something is.

E

um I can't find a detail of something well how how many branches something has it's um diffusion, limited aggregate type stuff, uh so it it is relevant to um looking at something and seeing whether it has a deterministic behavior.

C

Yeah we yeah, we haven't uh talked too much about chaos and especially diffusion limited aggregation. uh It's probably something we should talk about, though, um and especially with the basal area paper. We'll talk about that in a bit. Maybe that's useful there too, but yeah thanks for those papers. We might go over a little bit more detail later on in the meeting um and then let's see what our.

C

Chat says here so my ux says hello, everyone, my laptop charger, isn't working so I'll, be attending this meeting on my phone I'll be able to share my screen today.

C

I think I got some of the stuff you put up on slack, though my okay, I backed it up just in case, so I I can present that I can share the screen and then you can see it and maybe talk about it. A little bit um is everyone still.

H

C

Okay, all right uh yeah. I just want to make sure I don't, I think dick dropped off, but I wasn't able to.

C

uh I don't know if everyone still can hear me or not. I just wanted to make sure.

H

C

Good okay, so why don't? Why doesn't wall start um yeah? I can hear you. Okay, thanks.

D

Yeah, okay! So, since last week like uh I doesn't have.

D

H

E

This thing is being.

E

C

Well, it sounds good uh yeah. We talked about that in last week's meeting about the name of the different apps. So I guess that sounds good to me. I mean like if we had like the name of this organism and then divo learn after it, like whatever it is dash. Diva, learn something like that.

C

And it seems like it's pretty consistent with uh the naming convention. um Yes, yeah mike said he's on his phone today. So let's see if I can get uh pull up what he was sharing so actually let me yeah.

F

I

Like the distance matrix, basically like the difference between each node, like even.

C

I

C

Yeah, let me see if I can present.

H

C

I

Yeah kind of interesting and even bradley had a paper which he sent me. So I've been trying to read and understand. What's going on in there it's kind of going over my head, but I'm trying.

C

Well, we actually will talk about it a little bit. So, let's see uh first thing I wanted to talk about. Was this uh so the diva worm, the divo learn umbrella, so it came up it between usual and myself. uh Asking the question like how do we name how? What is the convention for everything?

C

So this is what I'm kind of envisioning here for the umbrella, so the main uh github organization is called devo, learn and then the software here, my ex uh pre-trained model would be divalern 0.9.1 and, of course, that version number would change with a new version or whatever.

C

But that's that would be that part and then over here we'd have the species specific model. So this is the thing that usually was talking about with like the species name or the uh organism. Name dash, devil, learn, dash or divalern heroku app whatever, and it would be just separate, heroku apps for each of them, so you'd go uh diva.

C

Learn would contain the links to those apps, which would be hosted on heroku app, because you can't host something like that on github, and so that would be the sort of the implementation there and you know they have to have their own name since they're, separate entities and then finally, the devo zoo, which is we haven't, talked too much about, but it would be sort of over on this side and also accessible from the divalern uh umbrella.

C

So when I say umbrella, some of you may not have been here last week. Let me see this is the tab.

C

So last week uh over the last week and a half, maybe we started a new github organization.

C

And that is called diva learn, so this is just in github. You can start an organization if you want to have like a face for your project or a face for your organization, that's distinct from what you're doing I have one for diva worm, but that's actually. I wanted to keep this a bit separate um just so people can find it and join the organization, and things like that. uh So this is a new organization.

C

um It has a couple of ready. These are the repos where ojawa and mayoka have been contributing. We have the diva learn, uh which is the pre-trade model. The c elegans diva learn, which includes a lot of what usual, has been working on the contribution guidelines, which are just basically how to contribute the media, which is a full. uh You know images and other things like that, and then this new one I made called data science demos, which is uh going to be for people to create demos uh in this data science analysis space.

C

So um my oak has been working on this network stuff which we'll talk about in a minute, and he can push there when he's ready um and but then also uh uh krishna katyal asked me if he could contribute uh something as well and it's. uh I can't remember exactly what it is. It's very data sciencey so I said sure uh just committed to this folder, so this is going to be the folder. If you have any demos you want to put up that are germane to analyzing data or something of that nature.

C

If it's like a tutorial in like a notebook or just like some methods or like some visualizations, this would be the place to push them and we'll organize this. You know later on by topic uh the networks- one was pretty obvious, but uh so when uh maya gets a chance to push to the networks, folder he's going to create um one of the things he's going to create.

C

Is uh he has his network simulation which we'll see in a minute um and it you can download the centroid positions and the uh distances between the centroid positions and all that would be in a csv file and then what you can do in github? Is you can upload a csv file?

C

If you just have the this raw csv format, you can upload that file to github and it'll. Actually I don't have it in this table. uh It'll render a table with the numbers in it. So it's as simple, as you know, if you have a cs something called csv, you can say something with a b c.

C

And then one two and then we can commit.

C

And then it creates this table that you can actually enter well, you can't interact with it, but you can download everything in csv format and it creates a nice table for people to browse online and then we can. We can use that as part.

I

Yeah, I I won't just upload that csv file I'll also upload a link to the executable collab notebook so that people can play around like with and it's all made on, divo learn, so they can just get the taste of the library and they can also play around with their own like video files, that's what I'm doing is with my own video file.

I

Maybe they have some other video. Maybe they want to use that, so they can do that. Okay, yeah! We can put these executable content, something that they.

A

C

Right so we have the yeah, so we would put the collab notebook here as well. Push that um and then you well. You could open up the collab notebook and I think it works best if you actually open it offline because in github it doesn't open directly all the time. So if people want to use it, you know it'll tell people how to uh open it up, but I like the way collab notebooks work. You know you can open them up and then it's connects directly to your um google drive and it's it's nice.

C

So we have a comment here. Jesse says: when you see an organization on github, it's something specific to join or contribute like normal github. Oh so this is like it's it's like normal github, except that on our end, it's an organization instead of an individual. So I think you can.

C

I think you can join the organization. You can't follow the organization. There's some rules like that. But it's basically like open worm is an organization. If people are familiar with the open worm github, you can't do certain things, but you can. Basically you can definitely contribute to it.

C

Anyone can contribute to it through a pull request, but you can also join as an editor if, if you're, you know, if you're contributing a lot like my oak and usual are contributors right now and you know, if you want to do you know, if you want to become a contributor, then you know it's. I can add you on, um but you know this is, if you're doing a lot of uh uploading of stuff and manipulating stuff, so I you know it. I would encourage you to check this out.

C

The diva learning organization see what everyone thinks. There's a lot of stuff we're going to keep adding things on uh over time, so this is meant to be very flexible. uh Then we can actually the reason. I want. Another reason why I wanted to create an organization for it, too, is because it's uh you know it's sort of a standalone thing, so you don't have to go into divo the diva worm github and look around and try to find things.

C

um So actually this was, uh I think, usually suggestion or my suggestion to put a to create an organization.

C

So that's the diva learn. That's we talked about the umbrella. We talked about the logos last week.

C

Now I think I'll go on to talk about the bacilluria papers, so this is kind of a shift uh we'll talk about the uh g-suck stuff more next week, um but I think my opinions will have done a great job on the evil learn uh organization and I think we'll probably launch that, like at the end of g-suck formally, so I think we've mentioned it a little bit to people, but it's not you know formally launched so we'll try to launch it, probably by the end of gsoc, uh and that means like we'll just announce it formally might do a blog post um and just kind of formally announce it to people so that the uh the uh open worm annual meetings coming up in september so probably get a good mention there.

C

So congratulations on that uh any other questions before we move on.

C

Oh actually, I did want to talk about my oaks network stuff before the best wario stuff, so uh I did get the I did download my stuff with uh the embryo networks. So embryo networks are something we've been doing in the group.

C

Like we did this stuff a couple years ago uh from the same data set, that mayor is working with, but he's doing a little uh he's actually doing it a little bit differently, but maybe not, let's see so he's generated these animated gifs. So this is a basic network.

C

uh These uh values along these two axes- I I imagine so the um the data set that he's working with um the epic data set they're.

I

Just like there's the dimensions of the image like this is just.

C

Yeah yeah the epic data set actually gives like position values in like sort of the space, that's undefined, so it just gives you these coordinates. um Well, one thing.

I

C

I

This isn't the.

E

I

Small dataset, which I found oh.

F

Okay online right, so this could use like I'll send that link all right later on. Maybe okay.

C

Yeah, so it well okay, so then maybe this doesn't apply. Maybe it does. But when we are using the epic data set, uh they give you these coordinates but they're not aligned across embryos just kind of like numbers.

C

So I did a normalization procedure for all of the embryos where you center the coordinate system on the middle of the embryo, and then it's kind of like where the p0 cell is the centroid of the p0 cell and then every you can do like a a z score and some other transform to get that number and then so, then you have a unified coordinate system and it's still something that I think I was referring to as embryo units.

C

Although it could be, you know you could. I think, map that to um an actual metric measurement. So but the the idea is, you have these centroids of cells and they're in a certain location in x, y or xyz space, and then they as the cells divide over time. You get new cells and they come up in the same coordinate system.

C

Then each of these cells are connected somehow by a distance, and so you have these cells that are connected by distance in this network forms. If you have this fully connected network and then you can evaluate the strength of these edges by you know some threshold you could use. uh Actually you can use a bunch of different criterion for these connections, like these edges could be weighted by their distance.

C

They could be weighted by maybe some hypothetical cell signaling yeah.

I

C

I

I

These lines are just like drawn okay, um so I'll just. I think it would be better.

C

If you show the next game.

C

So I mean this yeah. This is based on just the distance between the cells, though I mean this line, is you could measure it out? Yeah.

C

So this is the gift, so this is you know this is basically what happens in the embryo. Where I mean this isn't a real embryo, it's like.

I

Okay, just to make it look.

F

It with a black background.

I

To make it more clean, this one gives it off.

D

Yeah, so this is yeah well what I was like by seeing this gif so.

E

D

So basically, what.

E

I'm getting is that he has.

D

Right now connected uh all the centroids, with all the other centroids like in a form of a bipartite graph. So rather than just like connecting it with group by using blue force.

D

I

Like they are basically, if I'm not wrong correct, like in the initial stages, they are like two like. There are two basic images that is the a b and the p yeah, which would not identify, which is 18, which is b, but it can identify. Where is like.

F

So this yeah this is.

I

C

Yeah yeah, so this is the distance matrix here and smooth. So you have all these sort of components for a network, um so I was going to it.

I

Just basically shows the distances between each node from the other.

C

I

If there are like n nodes, if there are end nodes, like you notice how the square like the image, the matrix gets larger and larger with time, because there are no nodes like if there are n nodes, there are n cross and like that is the size of the matrix. So that is how it goes.

A

I

And I just tried to smooth it to look for some like broader patterns, but it wasn't much of a success. Yeah.

C

Well, I I actually.

H

C

I get I and now I took some slides out of some presentations that I've done on this, so I can send you those presentations if that's something you want to explore further, but uh so this work with embryo networks. This is something that uh dick and I and some of the other people from the um you know. I've been around the group longer have worked on. uh I presented on this a couple times. Actually this kind of work.

C

So the idea behind embryo networks is that we have these cells that you know divide and they form this embryo. But you can characterize it using a series of sort of edges. You can represent it as a graph, so you can have the centroids of cells, be these nodes and then their distance, the edges and create these graphs, and what you want to have is this sort of five dimensional representation, where you have three dimensions of space, one dimension of time and then this other uh symbol, which is some context. Q.

C

uh You know in this case it's I so that's uh the distance along the uh the ap axis of the uh lineage tree, but you can use other things for I uh you can use like uh some sort of you know. If you're interested in signaling, like an interacton between the cells, you can use some information and use. I is the parameter to represent that or are there other things you might be interested in?

C

uh This is a network that was made of the c elegans embryo from the extracted um epic data set, and so this is very much what they call hairball in network science. So it's just like nodes that are proliferative and they make these nice pictures, but there's got to be some way to you know sort of analyze what that means um that but network science people use these hairballs to sort of show the complexity of something, and so this is like all the time points overlaid upon one another.

C

So this is kind of like where you have sort of a lineage tree emerging, but also the distance is between cells and it forms this hair ball. This is a sort of a subset of nodes shown in a three-dimensional space. You can get a sense of like what it looks like when it's not hairball when you have this local neighborhood and then um you know you can use this. This is the paper I sent to um mayoque and this is of course, a paper that we have up on the website.

C

um I had a question on this. Actually, could you scroll up yeah.

I

Yeah, just a bit yeah.

C

Yeah, just this.

I

Yeah, I can see a 3d graph to the left, so how did you guys.

G

Annotate that graph like how did you guys know that this cell and this centroid.

C

Is like it belongs to this.

I

H

How did you guys determine that? Well,.

E

Is there some kind of.

I

Like is there some specific approach like.

C

Well, they can be.

I

C

In any way, no the epic data set comes in two flavors, so one is the images which allows you to extract cells, but they're not annotated, and then the other one is a uh a set of files that has like uh the distances and the names. So like that's, not an image file, that's like a more of a csv file, type thing and so that they have those data that you can get.

C

uh But you can also that's not image data, so you know it's just based on their uh like cell tracking that they have, and so that's that's where we got that from um you know. Each cell has an identity in c elegans anyways, and so you have the name of the cell, and then you have the position and then you have to average over the embryos that they give you, and it gives you this position. You know you can estimate a position for you know any stage.

C

uh So that's that's. They have those data and that's separate from what you're working on. um But of course, you know using say something like machine learning versus just the sort of cell tracking that they're using is going to get.

I

Just like this.

I

C

Yeah, so this is the yeah. This is the interactome and then um that's just kind of more of the same thing. So then there's this question, though that exists, which is when you get these networks. You know they look like hair balls, but they actually have uh structure in them. So you were showing those matrices and uh we have.

C

We can take these networks and create matrices like you did, but we can also analyze those matrices in ways that tell us what the structure of the network is, and so you can see here from this is from a paper on the sensory motor, connectome and zebrafish.

C

They, you have different patterns across this matrix, so you know you have something called a small world network where you have um they define in the literature.

C

I don't want to get too much into what these exactly are, but you can see in the matrix there's a certain pattern where most of your they basically threshold the connections by some criterion, and then they find that there's this diagonal sort of behavior a scale-free network, is sort of where there's clustering, but not diagonally. It's maybe along the edges here and then the random condition where there's just significant or edges above threshold across the matrix. So this is something you can do as well.

C

You can do these sorts of tests to ask what kind of a network is it, and so this was done with the c elegans data versus a random embryo and it's definitely significantly different from a random embryo. We don't really know.

C

We know in the c elegans connectome that it's a sort of has a small world aspect, but in terms of the entire embryo we don't know and again, these embryo networks are somewhat different than the connectome in the brain, because it's we're using a different criterion, we're saying it's just distance based and um but there's definitely information in here, um and so you know you can look at it like this. You can uh look at it in terms of like sort of the anatomy, so you know in a c elegans.

C

This is uh c elegans right before it hatches out of the egg. You can characterize the cells kind of in this spiral network. uh You can also take a mouse blastocyst, and this was based on a mouse data set that uh you know if you take the cells and you use the same criterion. You can map essentially approximate this uh phenotype here, and so you know again, it's it just shows you like aspects of the phenotype, but there is probably information in there.

C

We talked about topological data analysis, uh a couple meetings back and that's kind of what we're doing here, we're not really doing it in the same way that they do it in the literature. But it's a similar idea. So you know you might have like um you know, motifs within these networks, that you know we're just like kind of an area within a bunch of nodes and their connections, and you can measure those you know those little uh geometric areas and they might have some significance to spatial organization and things like that.

C

So we don't really know what this kind of network is. But you know there's we, you know there's something there, and I think this would help in trying to uncover what that is exactly because what we can do with the images is. We can like uh scale up the analysis and we can, you know just kind of generate topologies, and you know we maybe even they're, not even labeled, but that's not really what's important.

C

What's important is that we have the centroids, we have their relative locations and we have some criterion to sort them on and we can generate topologies that way, and so we can actually look at some of these questions about what kind of a network it is, and so it says a lot to do with cell division and where the cells go again like with the epic data set that we had.

C

The limitation was, is that you have the cells are kind of averaged in their kind of final location in development, and then they divide again and they move so they're moving around, but they stop at a certain point for a while and then they divide again, so they might migrate a little bit and then stay somewhere for a while and then divide again and form eventually some anatomical structure in adulthood um with c elegans. You know you have the labels, but the labels are only part of the story.

C

The real story is where the cells are going. So when we talk about like you know uh how the cell sort of changes it or how the embryo changes its shape, part of you know during development. Part of that is these cells moving around differentiating and positioning themselves with respect, maybe to other cells, or you know in arraying themselves into structures. We don't really know exactly how that happens, but we can look at like the patterns that they form that might tell us something.

C

So um that's I mean I have those slides if you're interested in looking at the slides, I can put a link in the chat.

C

You can look it over or if you have ideas about that, you can. Let me know um also. We have presentations which are on the website um and we you know we can go over that.

C

um So that's that's that and then the uh basil area behavior papers. uh I want to go over those quickly, so this is switching gears a little bit to basil area. We haven't talked about that in a while. We have the digital basal area project, which was something that started in 2019. This was the thing that resulted in the uh bourgeois was involved in that, and so was dick deep learning uh analysis of bacheloris. So you have these.

C

um You have these diatoms, which are marine microorganisms and then, in this specific type of diatom, they live in colonies and the colonies are sort of have this coordinated movement. uh What we did in the first paper was to analyze these phenotypes.

C

uh You know: try to figure out the shape and sort of the shape parameters and build like these, maybe these digital models of these colonies. So you have all these cells and their shape, and then we also did bio mechanics, which we don't have images of here. I didn't really up this with respect to that work, but um that paper is available.

C

uh This was uh thomas harbick, who is a diatomist in germany, he's been doing a lot of stuff with diatoms and looking at their movement. So he's been doing this thing this. um He did the analysis, the biomechanical analysis of this. So that's what uh the digital basilary is remind us. What it is the next step from that, though, is to come, commit this from a behavioral standpoint, and so this is what these two papers are and jesse expressed some interest about this in the last week.

C

So uh there are two papers working papers right now. One is uh the. This is the one that was submitted as a proposal, the psychophysical, the psychophysical world of motile, diatom, bacillary or paradoxa.

C

This should be italicized, but that's just the formatting here. um So psychophysics is an area of like psychology, where you know there's this measurement of different things in the environment, like physical things in the environment, the nervous system sort of takes in information and operates on in a waffle manner.

C

um Beyond that, I don't want to get too much into what psychophysics is right now, but the idea is that there's a psychophysics of behavior- and this is something that you know. This is just something you see in humans. Well, it turns out that you see it in a lot of different nervous systems, but you also see it perhaps in bacteria, at a citation of a paper where they looked at this in bacteria and in slime, molds things like habituation.

C

So this this is psychophysics. This is something that is probably most germane to like sort of unifying the behavior of something like vassal area. It doesn't have a nervous system with these sorts of ideas, and so this paper is about going through and looking at, maybe what a bacilluria colony does in terms of its movement that mimics cognition or mimics like a nervous system. So bacillaria are these: it's this colony of microorganisms, the different cells. They move relative to one another, they're actin filaments, that facilitate this movement.

C

So it's a lot like what muscle does in animals. You know in animals like you, know, fishes and humans and all sorts of um more complex organisms. You could say, um but they you know they have the they may have this set of rules. They may actually act like a nervous system in some ways, and so there are these different uh models of behavioral regulation.

C

We can test, we can think of it, maybe as more of a hebbian mechanism, which is something that, if you've heard of the dictum where together fire together, that's what that refers to. So it doesn't require neurons, but it's like it requires coupled systems uh and then there's pseudo-intelligence.

C

So maybe this is sort of like you know, a null hypothesis, maybe there's just something that looks like behavior, that isn't really behavior, and so we've considered that as well, and so there's this literature that we're drawing from as well on a neural cognition.

C

So this paper by michael levin and belluska on having no head cognition throughout biological systems. They talk in this paper and there are a couple others uh about this idea of non-neuronal cognition. That cells actually produce cognitive behaviors.

C

That are, you, know, not generated by a brain, but they look like they're generated vibrate, and so um this table kind of goes through that a little bit table needs to be worked out a bit more.

C

But um you have this instance of behavior, which is, if you look at something, does it look neuronal or non-normal, and then it's generated by some process underlying that behavior, so, but then that can also be neuronal or non-normal, and so bacilluary chains fit somewhere in here. So, for example, a non-neuronal looking instance of behavior that's generated by a non-normal process is the steam engine.

C

It's just an example, um something that's generated by neuronal processes, but that also looks neuronal could range from like the brain to something like an ant colony and so um that you know their differences and how those look but qualitatively that's their category uh cell migration is non-rural, but it's uh maybe have this wrong. uh This is a cell migration has a certain you know: there's, no, not the cell doesn't have a brain, but the instance of behavior is maybe I have this in the wrong category.

C

But uh this, like it says, table, needs work, but the basic idea is that you have instances of behavior and they're generated by some process and you can match those up, and so this is this paper. You know this is just a proposal. So, there's a lot of work to be done in this.

C

um It's basically about interpreting the um behavior of the basil area to see if it you know what it fits into it is it maybe it looks intelligent, but is it really intelligent and we can consider those things.

F

C

Paper, uh the other one that I have, that kind of been working back and forth on is this collective pattern generators.

C

So this is something that isn't really tied to any obligation, but this is something that um so there are these central pattern, generators that exist in say even like c elegans, nervous system and so c elegans nervous system uh has these. You know neurons that fire in a tight circuit, that's called the central pattern, generator they fire rhythmically and they produce these patterns, like maybe like tail movement or body oscillations, something that's rhythmic, and so that's something that we see in vascularity as well, except that it doesn't isn't generated by the central pattern generator.

C

So the idea that uh is presented here is this idea of collective pattern generators, and so this is something that you get from sort of. uh You know electrical activity or even like something like actin filaments, doing their thing, but they're sort of integrated into a system where there's a collective behavior that's generated, and so we know that collective behaviors can look quite intelligent, they're coordinated, but they're coordinated at a higher level. So we never really understand too much about that.

C

We can talk about internal architecture, so we just talked about how vascularity doesn't have a brain, but it's generating these behaviors and then how to model this. So this can be modeled using a series of sine waves, other types of functions to represent the movement of the cells- and this goes back to the data that was collected earlier on vessel area and also how to map these measurements to the data set. So we can do a simulation of the movement and we can look back to the data and we can from that.

C

We can get some information about how this is this uh sort of collective pattern generation works. We can also look to other types of central pattern generation.

C

So there's a data set um cpgs in a stick insect which actually is a stick insect is an insect with about six legs and it walks along uh and when it moves its legs. It generates these central. The movement is generated by central pattern, generators but they're coordinated because they have six legs, and so um I don't really have much to share on that data set. But that's there and they consider in that paper how those cpgs are coordinated. They don't use the term collective pattern generator, but it's this is.

C

I think this would lead into this idea nicely. So you have these simulations of sine waves and this basilaria colonies are basically a bunch of overlapping sine waves in time, so they exist different it with relation in one another in different phases, and you can also them with noise and you can evaluate the behavior and then that that would be that. I don't really know you know this isn't really tied to any sort of uh book chapter or anything. But that's another thing this.

C

This might end up becoming like one paper or the copgs might be included in the first paper, not really sure that'll work, but I just wanted to give people. Is it a gate and a like gate in a mammalian brain?

C

Do you mean central pattern generation, or I think I don't know, I think gate does have like a central pattern generator. So um the you know they're like there's a whole literature on this. I know for sure c. Elegans has them the bona fide and a lot of insects have a central pattern generator. I think people have hypothesized that um that animal movement, animal walking is generated by a cpg.

C

um So I mean you know and then in like humans, it's modified by all sorts of other things. So um jesse said thanks for the overview you're welcome, so um I go back to the um so. I came to me a couple weeks ago and asked me about building a bibliography of all the work that diva worm has done, and so I said yes, that's a good idea, so we have, I sent him a bunch of papers.

C

I sent him a list of the papers with presentations and posters and then also some of the papers that have been presented in this.

C

C

Just anything that we've hit on that looks of interest to the group, so this is something that is ongoing. If you yeah I'll, be curious to see the bib yeah yeah, so that's good. um If you're interested you can get in touch with dick or myself jesse seems interested in it.

C

It's just something that I think is good as an organizational tool and we can link it into the evil learn stuff as well, because this is you know, medieval learn stuff, on the one hand, is like a lot of image, processing and everything, but we also, as we pointed out today, we have a lot of things that we've been doing, that kind of fit into some of the exploration of the data.

C

So you know it can generate a model like a network, but you know, maybe people have already done this type of work in the group or you know outside the group. Even we might want to have some sort of collective knowledge about that, and it's a little less than a google search. It's a little more concentrated yeah. So may start this week, okay, yeah! So let's, let's keep yeah.

C

Let's keep revisiting that um that's good um susan said the cerebellum creates gator walking motion, yes, cerebellum has a architecture of like, uh like oh we've, talked about this in another group, where it's like a internal model of the environment, and it generates motion. So the the cerebellum isn't a central pattern generator, but it does modify or it does mediate.

C

uh Human movement um is thinking you know more along the lines too, of like fish fishes swimming things like things that, like your heart rate, things that create a rhythm that require sort of a rhythmic pattern to be there. You know, so it's an interesting topic um yeah. This is so this is the endnote basic. So dick is going to create this an endnote.

C

So endnote is a bibliography software. People are, uh I know, a lot of the students here are like always doing papers to read- and I know susan probably knows this and jesse might know about this.

C

Dick definitely does, but you know you have to have some way of like organizing it and um you know outputting the preferences in a format that you can, like you know, manipulate, because if you get to the point where you're dealing with like hundreds of references, you know this is where a bibliography comes in handy and so dick has, I think, a bibliography of hundreds of thousands of papers.

C

He has a master bibliography and then he sliced it down into different categories, which you can do with a bibliography software. So he's going to build this yeah. The commercial version is in note9 he's going to build this an endnote and and notes. You know. I think you can also use the endnote files like in things like mendeley, I'm not really sure the cross um yeah the cross, uh but that it's basically it's it's very good.

C

You know you'll be able to it'll be compatible with endnote, and then we can do things with the references, so that'll be a good thing to have so I know this is the top of the hour. If you can stay, if you can't stay, that's fine, but if you can stay for a bit I'd like to go a little bit into them, I don't have it open anymore.

C

uh I was going to talk a little bit about the oh.

C

I was going to talk a little bit about the chaos measurements, um so jesse says is diva, learn the correct use of capitalization, or is that not yet set? um I think, that's probably a good style uh stylistic presentation with the capital l just because we have the diva worm with the capital w.

C

But I don't you know, I don't know if it's uh I think, if you don't use the stylized l people understand what it is. I mean you know it's, uh but I think we'll probably stick with this stylization because it's I don't know it's like two different things and you're bringing them together.

C

What do you think which one.

D

So I think, like whatever library name, is all in small caps. So if you are using it in a python version on the terminal, so it is all in small cap, but yes like uh for the logo and the name of the rapper or I think like uh something like. uh I don't know about that or us much.

C

Yeah, okay, so, let's see let me stop presenting for a minute. I'm gonna pull up those papers again from the uh I'll pull them up from the history.

C

C

C

So now I can pull back up the screen share here.

C

So susan, I think last week sent me a couple of papers on uh on chaos and.

C

This is the now: let's work back to that.

C

So there here it.

C

I

C

There we go, and so these papers are good. We have this thing ongoing with the measurement uh project or the idea of biological measurement. I know that a couple of us expressed interest in this and it's a sort of a sub interest area where sort of like biological measurement- and uh you know formulating things. I can't remember what the name of the thing is on the task board, but that's not really important right now.

C

um If we get into the papers.

C

She sent three papers on this if I can get into them.

C

um And they basically lay out like susan was saying earlier: it basically weighs out this model of like measurements of dynamical systems. uh It also lays out this way to distinguish between sort of deterministic, behavior and chaotic behavior so like if something is deterministic versus chaos.

C

There's this whole literature about the edge of chaos which has to do with like systems that exhibit sort of fluctuating behavior. That is um sometimes it's almost random, but it's not quite random.

C

I think if you uh george mikolovsky, did a presentation on something called complexification, which is this idea of like how a system will can go from like complete determinism to complete randomness, but there's a sort of phase in the middle, where you have yeah, where you have this sort of uh potential for fluctuation, and maybe it enables you know, organization and things like that to emerge.

C

So white noise, brown, noise, pink noise yeah. So there are different types of things that you know. There's this role of noise in systems for one thing um and systems can be noisy in different ways, and so we think of white noise, which is like uh the static that you see on. Maybe like a television screen or a computer screen and that kind of noise, we think it was just kind of synonymous with randomness, but there are other types of noise like brown, noise and pink noise that have like this.

C

They call one over f structure, which is um it's an exponential distribution which means it's. You know not a normal distribution, it has events that occur in time that are larger than other events, so you know it can right. It's broad noise is like brownian motion, so it can do things like you know, so it's not just randomness, there's some structure to it. So, if you think of brownian motion, that's brown, noise, pink noise is like music and then I think you know I don't know what yeah there.

C

If you look online and you look up like colored noise, you can get a good treatment of the different types of noise that there are because.

C

I'm having trouble opening this, I don't know why. Let's see.

C

I'm just trying to open up these papers, it's exceedingly hard to get into the boulder for some reason. But um well maybe we won't present the papers, but we'll talk about it a little bit. um So that's the idea that there's these chaotic measurements that you can you know you usually have like a dynamical system or time series and you look at the time series and analyze it in different ways. um I know that, like in jesse's probably familiar with this from narrow match, but they looked at dynamical systems, analysis there and they're.

C

You know, I don't know if they're using chaos, it's a bit old-fashioned, maybe for the brain, but uh there's a guy walter friedman published a lot of this like in the 70s and 80s on uh on. You know, chaos in the brain and things like that, so they did. You know different measurements at the time series, synchronization measurements and other types of measurements, and they claimed you know they could observe things in the brain. Like you know, synchronization and chaos.

C

You know at the edge of chaos behavior and things like that, I'm still trying to get into these papers. I can't seem to do it, I'm not sure why.

C

Let me see um not letting me get in oh well. Well, it was a good idea, um but anyways I just wanted to meet yeah dr white told kiss kinsmer from the university of manitoba as an expert. So that's your local expert on this stuff. So yeah look up this name here.

C

um There are other ways we can like. You know, learn about it um they're, but like some people, you know, but in the 70s and 80s, maybe in the 90s people were really interested in chaos, theory and- um and there was a lot of stuff having to do with like visualizations of complex systems where they visualize these attractors or, like you know, diagrams of trajectories that were like you know, loops and things like that, and there are some actually quite impressive, uh space-based diagrams that they called them and they were very beautiful.

C

But you know then actually mapping them to a complex system was another matter, and so they do actually use chaos as a measurement for things. Susan pointed out pointed this out in the chat.

C

It's a good measurement tool um and it's definitely something that has been developed. I think it's like one of those things where there was like a peak of interest when it was like beautiful pictures and then when it became more, you know it became a tool.

C

People were like. Well, it's not that interesting. I mean there are people who continued on with the research, but.

C

um Continued on um yeah, I'm not getting this paper. I don't know why this won't. Let me in but.

C

Google drive is rolling out my friend today for some reason.

C

Well, anyways, I yeah, I can't get into the papers right now, but I think I think, if you're interested in this area, we should talk about it more in depth. In the following weeks: yeah strange attractors and the butterfly affecting weather so yeah, the so the the diagrams that I was talking about. Are these phase diagrams?

C

uh If you know anything about strange attractors um there, you know you can look up strange attractors on google and you'll see all these pictures of. Like you know, attractors two attractor points three attractor points, these fancy orbits.

C

Basically, the idea is that the system has stable states in this phase space and these uh orbits are the system moving around those space, those points, and so you know you might have a stable point where you know the system is sort of stable, but the system also exhibits a lot of variation around that point so and then the butterfly effect, if you've ever heard of that is where you have some small, introduce some small perturbation at one point in time and it has a multiplicative effect to things later in time and that's a very popular conception of this sort of thing, the butterfly effect where, if you, you know stamp out a butterfly here in in my location, I caused some sort of weather, anomaly and another continent, because there are a chain of you know, causal things that happen that lead to this event happening or not happening, and so the systems have a sensitivity to initial conditions.

C

So if they're, you know some, and if you change the initial condition, you can change the system's behavior, quite appreciably. um This limited diffusion aggregation is another interesting concept um which I don't know. If we can do without any sort of visual aids, but um it's basically this idea- and I think we've talked about it in previous meetings, where you have like the space and you just you start, maybe with the single um clump of something- and you have a bunch of things migrating across the field right.

C

It's like a field of like things that you know. So you have this thing in the middle of the field, and you have these things that come from the external part of the field and come up migrate across and they do this at random and eventually some of these hit the object in the middle where they hit each other and they form a cluster, and over time these clusters will grow because things bump into them.

C

This is just a random process and you end up with these structures that emerge from basically just a bunch of random motion, so things are moving randomly across this field. They collide, they come together, they become stationary or they start to move in a different direction.

C

They yeah snowflakes are a good example. They form these patterns that you wouldn't get otherwise, and um I don't know we I think we talked about on the group. We I think I've had some. um I think I had a couple. I have like a bag of simulations that I have um that I use for instruction and I didn't bring it with me, but I think diffusion, limited aggregation is one of those where you have this and it's it's it's better to look at it.

C

When you have this, um you know when it's a visualization than what I'm describing it, because what I'm describing is not like that impressive. But if you see it in action it looks a lot more impressive, so maybe we can put together. I don't know I was thinking. Maybe we could put together some simulations and we have some of those in the education, um the dwow education um repository on github.

C

We have some simulations already based on sort of, like you know, sort of more germaned classic embryogenesis um like a lot of pattern formation uh simulations, but we don't really have anything in diffusion, limited aggregation and we don't have anything on like chaos. We have a little bit on like attractors, but um that that part of the repository was just pieced together. So it's you know it's one of those things where I think we could maybe use our github repositories to build up some more information about that.

C

Maybe apply it more directly to like some developmental process and we could use it either for education or we could use it for this measurement that we were talking about, because the measurement area is pretty open.

C

um I'm not quite sure I don't know if I have it I'll see. Now it's letting me into these papers.

C

Okay, well anyways, it's too late for that I couldn't get into the papers. It was. Google drive, wasn't letting me in. So that's fine! um So, let's see we have one more thing in the chat: yeah snowflakes. That was the one we talked about. Okay, so I think that's good. I think we can maybe talk about that in future meetings about the chaos papers. I wanted to kind of go through them, but I didn't really have time.

C

You know I couldn't get into the folder, but I think we can follow up on that, maybe with something in this measurement area or maybe in the education area. I think the measurement area is interesting, because I mean what I think the goal we want to do there is to try to build up. Maybe some measurements measurement techniques in that area.

C

Maybe like have you know some code, perhaps to do some analyses um or you know, even if it's collected from somewhere else, we can kind of host it locally and say: okay, this is a possible analysis for our this type of data.

C

um You know we could have simulations, so we could have a lot of things there potentially and then jesse, and I talked about that once um but yeah there are things we could do. Yes, I can try to find more information on this for you yeah. Maybe we can find some more information on specific measurements.

C

If there's something we want to follow up on on those papers or doesn't even have to be related to those papers, it could be like something else entirely.

C

um Or just so, you know anything that looks like it's developmentally relevant. We could use it.

C

Okay, well, thank you, everyone for attending. Is there anything else anyone wanted to say before we go.

C

Or okay: well, thank you for attending and next week we'll have another meeting yeah from okay good uh next week, we'll have another meeting. If you have anything you want to contribute, let me know we'll have updates, as we did this week, stay safe, everyone, okay, you have a picture of two doas. Well, send them to me and I can bring them up next week. Hopefully, it'll google drive will cooperate.

C