 Well, thank you all for coming. My name is Morgan, and I'm just representing Neurofidora today. And, yeah, we've got, well, would you like to introduce yourselves? Introduce yourselves, the two of you. Everybody knows me, okay, so I'm Sbyshek. I work on packaging of Neuroscience and science related stuff, and I am creating a festival. My name is Luis Vasan, I am a member of the Neurofidora Seed, I am a packaging maintainer, member of Federal Ambassador, and a lot of other teams, I don't remember the other teams now. And sorry, I'm Morgan Huff, and I'm kind of representing the neuroimaging tools in Neurofidora. And Ankur Sinha, who couldn't be here, is more representing the computational neuroscience tools, and has been, you know, the tour de force in terms of packaging. Anyway, let me, these are Ankur's slides, so bear with me if I'm reading them with you. So, yeah, we started this, certainly more, I was more interested in brain imaging tools, and here we've got a number of examples that we've got tools currently in, not in release, but in queue for, you know, looking at the neuromophology. There's lots of different kinds of neurons that you can model. And here we've got a bunch of different brains, and comparative anatomy is another really interesting area. You know, a lot of the studies that we look at are, you know, probably started with mice or rats, maybe even monkeys. And of course then we try and translate those into human studies. There's a great repository at the Institute Pasteur for, you can get MRI data on all these kinds of animals. Naked mole rats, dolphins. So, as we all probably hear about the thousands of connections between neurons, and these kinds of models that we use, you know, both in computational neuroscience as well as, you know, in machine learning, and kind of reinforcement learning, which we'll talk about a little bit later. And what's going on at these synapses really underlies our best understanding of learning today. So we study a lot of different kind of levels and aspects of neuroscience, the physiology and brain functions, how it's structured in a neuroanatomy, and then how we know that's, at least how the physiology is broken down in terms of pharmacology and biochemistry. And we can hopefully be doing a better job in splitting the tools up and showing you which tools are used for kind of which aspects of this. And now, of course, what we care about are these kind of computational models, both in terms of how the brain's doing something, and of course building up our own computational model on the computer of that process as well. And these all relate to our concepts of the kind of behavioral units in the behavioral analysis and the cognitive analysis. And I myself apply this in psychiatry and it's just one of the diseases. I've also worked on multiple sclerosis, rheumatoid arthritis, different clinical areas that you wouldn't necessarily think have kind of a brain or neuro component to them. But neuroinflammation and neuroimmunology are two really important areas of development these days. The human brain project, and I've brought a few brochures that I brought from Geneva. This is more of the blue brain project, but the human brain project, you'll see some brain-inspired computing. And there's brain scales in University of Manchester and then there's another group in Heidelberg. Neuromorphic computing, something that we don't have any of the tools currently packaged, but that's something where these are all areas that we could use more packages for. And I certainly started in philosophy and consciousness studies are also a very hot area, at least in neuroimaging. We look at locked-in syndrome and people in comas in terms of if they've lost the ability to actually communicate, can you show a brain response to show that there is consciousness and kind of the philosophical and ethical implications of that. At least in the case of Cambridge and Adrian Owen, who's now moved to University of Western Ontario, they are looking to get consent from patients in terms of these people need a rather dangerous surgery. Can they ethically consent to the surgery? So some of the research, well, here's just a general workflow that you probably see in any science lab. And we're talking about Neurophrodor's Day, but we really want to see more cross-pollination between all the different SIGs. And this kind of, we have theory leading into modeling, leading to experimental work, leading to data analysis. And of course in some of these different tools are being used. This is probably better directed graph representation of the relationship between these areas. Okay, so tools of the trade. This is maybe a little more my area in terms of here are the kinds of systems and the kinds of modalities that we collect. So EEG and electrocortograms, those are the intracranial recordings, as well as single and multi neuron recordings that you might do on just slices of tissue or extracted nervous systems. And those are going to give you a bunch of time series. These days it's going to be multiple time series. The CT and the MRI, those are your more well-known 3D volumetric techniques that are giving you that great representation of neuroanatomy. DOI in this case is diffusion optical imaging. So we can use near infrared light as a means of actually capturing, capturing blood oxygenation. So fMRI, which is the MRI technique for looking at blood oxygenation, can actually be much more cost effectively replaced with just a near infrared sensor and it will shine light right through your skull. So we do have some packages for near infrared imaging. And then MEG is actually another EEG package or modality, just uses much more expensive sensors. PET is the last one there as being a more real nuclear radiation technique where you're going to ingest something that's going to be emitting some high energy particles. But PET is a terrific way of capturing a lot more of the biochemistry and organic chemistry depending on your sophistication and making tracers. So a lot of the neuroreceptor mapping that's been done in humans has been using PET techniques. So of course in data analysis, the data analysis packages that we really want to see more of are going to be the ones that you're probably pretty familiar with in computational Python. A lot of the computational Python packages were actually developed by fMRI researchers. So I don't know if anybody is familiar with Seaborn, but that's a pretty common one for all sorts of disciplines. He was an fMRI researcher at Stanford. He's now moved up to the University of Washington. As well as, you know, our packages and then the kind of statistics behind things like MCMC, so JAGS and STAN and other tools that you'll see for what I would call statistics. And then, you know, what we're going to hopefully talk about a little bit at the end is some of the more traditional machine learning or deep learning tools and making sure those are packaged too, because they're certainly, we see a lot of those being used in medical imaging for both nonlinear registration and segmentation. So it's really improved those areas. So we'll see a lot more dependencies of big packages on those tools. And then simulators. You know, we really, there's a great paper in Nature just recently about the real scientific utility of simulations. And they got a lot of PIs from computational neuroscience labs to contribute to that article. And that's why we were presenting at computational neuroscience in Barcelona this year, where we had a poster. So, yeah, the tools, the other tools, the trades are the ones for, you know, academic writing, blogging, podcasting, video making. Yeah, we see a lot of scientists these days are, you know, their own social media agents, and it contributes a lot to their labs. And yeah, so we're certainly looking to see those tools in. I'd certainly like more tech live support, but and under collaborative tools and utilities. So I think this would be familiar for everyone who's at a Fedora flock conference. Everyone should have freedom to share, study, modify scientific material. And of course, you know, you don't need a PhD to be a scientist. You don't have to publish in, you know, Nature or Science to be a scientist, but you do. You need the tools and you need the scientific method is basically what requirements of being a scientist. And that is really what's behind, you know, us being here at a Fedora conference in terms of this applies to software, which are really tools of the trade here. So free open science implicitly includes relies heavily on free and open software. And here's a particular, you know, we could pull so many, so many great references in the last couple of years. And, you know, I wouldn't want to leave out the open access to journals because, you know, that has been as important as anything else. I don't know how many times where it's like if I'm not connected to my VPN for something where I link, you know, somebody's got an actual journal article talking about open science. And I go to the, I hit a paywall saying, like, please give me $35 to read this article. And sometimes it's, you know, $135. So the, yeah, open access to journals is as important as I'd say the software. And, you know, it's one thing to say, yeah, it's not just the scientific method that you need, but it's also what are the open problems in science today and you get that from reading the literature, you know. So just a little plug there for open access journals. So what can we do at Fedora to help? Well, neuroscience, I mean, certainly one of the things I love about neuroscience and psychiatry is its multidisciplinary nature. You know, we are constantly working with people in different areas. It's really, you know, interdisciplinary teams are the rule pretty much. And, you know, probably should have statisticians attached to this list. But, you know, theoretical mathematicians, you know, certainly physicists and chemists and biochemists and linking in, you know, metabolism, immunology and other areas. And then, you know, understanding where the cognitive scientists or psychologists are dealing with too. And every one of these groups requires another representative, which is, you know, software engineering, really. And, you know, whether that's, I mean, every lab that I've been a part of has had, or every great lab that I've had the chance to be with, you know, has had a really full-time dedicated IT staff that has been why that is. That institution is so productive. You know, people coming in, not having to worry about setting up computers, not having to worry about software conflicts and things like that, but just being able to get to work is another mark of, you know, a well-staffed lab. So, too often, you know, we see this in terms of dealing with upstream, you know, what you're really talking to is probably a single developer working alone, or small development teams where they all know each other, you know, access to, you know, particular hardware and resources, although that's, I think that's changing more now with cloud computing. And certainly heterogeneous code quality, let's call it that, and limited use of established best practices, because, you know, again, usually the developers are also the scientists, or more typically, they're the grad students that are developing the algorithms. So, you know, again, these, I think certainly things are better than they have been, but, you know, there's not really an established best practices that they're widely applied. How much testing goes on is another question. And, you know, the other thing being once you want to put things in the door and you want to get other labs using code, are they using that code for the same reasons that that person developed the code? Because they might have tests, but those tests are going to be pretty specific to that particular paper or that particular problem. And, you know, that's what leads to the maintenance and, you know, the interesting project life cycles that you see in terms of, you know, that person's usually working on that code until that paper gets out, and then, you know, their priorities are going to have to change. So, yeah, depending on the lab, you're going to have a lot of different dependency chains, and some of those they get to pick, some of those they don't. And, yeah, documentation obviously is another big hole. So, we've got a lot of projects where we've got, we've got a lot of code that we would like packaged, but a lot of those projects are lacking documentation. So, yeah, the, you know, the community development, I wouldn't say that the know-how is lacking, I'd say that the ability to get people's time is lacking. You know, everybody that I know that has packaged codes to release is super interested in supporting it, but, you know, they just don't have it under their control to give you their time. So, what we want to do, we want to, you know, try and reduce the time wasted, the effort installing, reinstalling, debugging, you know, working out dependency problems, and try and get people more aware of, you know, how these things could be resolved, hopefully in the packaging stage, hopefully in our own testing, and what we bring in terms of having a process in Fedora that we can, yeah. So, making sure that, yeah, the test suites are actually run, and things are reported upstream. And certainly that's something you see where I know I saw someone saying that their software was downloaded 1400 times, but, you know, if it's buggy, then, you know, that doesn't really matter, and how many of those people are going to actually tell the person, well, I, you know, tried to run it, but never could, you know. So, just the downloads aren't necessarily the best metric that people should be using to say, yeah, my code's getting used. No, not really. Anyway, what we're trying to do is be a better liaison between the upstream and users. You know, bring those best practices to a development process in terms of what is our scope as packaging. And certainly we've got potentially better infrastructure for people to look across, you know, across different processors and across particular versions. And what we want to do is really help grow the community and, you know, and mind share. Learning from one another, training as we work, and disseminating information to end users. Okay. So, yeah, so why do we start NERF-Adora? Well, it was really to provide this integrated free software platform for neuroscientists and brain imagers. You know, and I would say, again, like across all the science sigs that we're going to have, we're going to have a lot of shared goals, too. And then really improve the standard to maintain maintenance of these tools, help users develop their own software development skills. This might be the first place that they're going to get some feedback because, you know, it's usually not their advisor's job. And, yeah, and I would say not just make these neuroscience tools accessible to non-specialists, but also to show how non-specialists can really contribute to a scientific software tool chain. So, and obviously to make Fedora the go-to distribution for neuroscience and hopefully for open science. So what we have been doing, so leveraging some community resources to this new domain and taking the community model of free open source to neuroscience research. So less than a year old in its second iteration, but it's really been great seeing it come alive. And we've got 15 active contributors, 10 package managers, package maintainers, five designers, newcomers. There's a core of us with neuroscience backgrounds, both computational and neuroimaging. And I'm certainly super interested to help others learning. I brought a few brochures of projects from neuroscience, computational neuroscience projects. And we've got 105 packages ready to install today, you know, even if you could try those out and make sure that you can't, they do run. You'd already be contributing. There's a whole lot more in the queue. And yeah, we just presented just in July. I was at CNS in Barcelona, which was, it was great. They had a really great day right before the meeting that was mostly covering software packages. And which tells you how important that is to the scientific process that this, you know, almost half the conference was just training people on software tools. So what do we want to do? Make more software available. I know that neuroimaging sides, we've got some big holes in terms of complete pipelines. And obviously improve the documentation. That's, there's a lot missing there. Once you are dealing with the end users and, you know, how much we can do in terms of rolling out some support, but definitely increase the community and convert, you know, a research user base into free and open source software contributors. And, you know, I know so many labs that really are doing things, but they are, you know, effectively they're siloed. And it's not for, you know, it's not that they don't want to contribute, but a lot of them just don't know how, you know, what are the best ways to get their work out there. So, and then of course convert free open source software contributor base into users. And that's something that we'll try and link some of the educational opportunities that are part of human brain project and things like that. To is around getting software engineers involved in scientific projects. So anything is just more for Dora really. Packaging, testing, containers, documentation, evangelism, marketing and design. And here's, here's our mailing lists. We're an IRC telegram, the docs and blog. And of course we're using all the normal Fedora tools for infrastructure. So, there's more science in Fedora than just, than just Nura. So, you know, I would love to learn more from people in astronomy, big data, machine learning, electronic lab sigs, medical and sci-tech. Some of these have, you know, been kind of sleeping for a while. Some of them might not really have any active developers. But I will say just, you know, checking a few key search terms last night. You know, that some really old sig pages come up if I just put in some simple terms. And, you know, I'd love to get those updated. And that, you know, there's a lot of common infrastructure in terms of scientific tools across disciplines that we could be sharing. And these days, the relationship between machine learning and neuroscience is, you know, becoming more and more interesting. And certainly I attend conferences on both, as well as specific conferences that look at the interaction of those. So at Neural PS this year, I know there's going to be AI neuroscience special workshops. Montreal, University of Montreal, McGill do an annual machine learning and AI neuroscience meeting that will be coming in October. And then for medical imaging, there's another great conference that is in China this year. And what you're seeing there in medical imaging and radiology is the overlap of deep learning. You know, deep learning is really taking over a lot of automated radiology and medical imaging projects. So neuroscience is not all about working on core research. You know, like I said, there's a lot of, well, there's a lot of best practices in software engineering that are really overlap with best practices in science, in terms of reproducibility, in terms of testing, in terms of documentation. And, yeah, glad to see that. Open source brain I'm not too familiar with, but, you know, this and human brain project, there are a lot of really accessible sites that I'm happy to point people to, to learn more. Open worm, again, I'm not sure how active this is, but, you know, there's a very simple C. elegans. It's like 128 neurons, I think, or something. Anyway, it's something that should be relatively easy to create a full bottom-up model. They're still working on it, but there's some great projects that relate to that. Here are some, this is something I'm a little more familiar with, although I don't know these particular people. But again, Montreal does an annual, and human brain mapping does annual science art, where people take various things. I thought one of the great, one of really interesting ones was somebody who programmed an MRI scanner so that its beeps and sounds actually played a tune. I thought it was also interesting in terms of making an MRI actually play a song. Different kind of art, but anyway. So, myth too, only researchers can do neuroscience. It's too hard, absolutely not. But it just, it requires, it's that same discipline, you know? And so, if you're familiar with good practices, best practices in software engineering, then, you know, in many ways, you're ready to do science. And so, this particular, he's got a little, yeah, spike time-dependent plasticity example here. You'll see, I guess the blue neuron is representing a perception of food and then another of smell, or is that the smell and then determining curry? Not sure, but what you're seeing there is that this is two neurons that fire together are having some plasticity. And these are getting modeled. Here's just an example of Nest, which is one of the things that we do have packaged currently. Nest is a great package for doing large simulations of a lot of neurons with some sort of connectivity that you've got in Matrix. And I don't think there's a picture with that, but anyway, why don't we thank you so much? Hello, everybody. I'm here just for two or three minutes. My name is Lumir Balhar. I am a Python developer that analyzes and Python maintainer from Redhead. And as Morgan said, a lot of six in Federa was inactive for quite a long time. For example, machine learning was inactive for the whole time it exists, basically. But I'm here to kind of revive it. And I need your help and I think that we are a lot of common machine learning that analyzes and neuroscience has been set. So the main goal of machine learning is to make Federa the best platform for machine learning and artificial intelligence developer. And we really want to do that, but we need something from you. And what we are looking for is we need a lot of information together because I am, that analyzes, I am machine learning developer, but I don't know anybody else who did it except two people on mailing list. Yeah, it's not that much. So if you want to share your experience with machine learning tools, libraries and software on Federa with us, it will be really appreciated. We need to know what doesn't work for you. We can help fix it, package new stuff, etc. What works for you? Also, we need to know some success stories. I mean, if you are using some machine learning that analyzes, etc. tool in Federa for a long time, let us know and we will create some blog post promoting that Federa is a good platform for machine learning and that analyzes. If you miss something in Federa, again, let us know and we will do our best to package it and to make it ready for you. As I said, some success stories and if you create an interesting project, anything related to machine learning and artificial intelligence, let us know, please use mailing list. I'll show you contact me in a while. What we can offer, we can help you with packaging, new software or also package maintenance. If you have a lot of packages you need to take care about and you don't have that much time, let us know. Also, if you are in neuroscience or in science at all, if you have any problem with machine learning, data analysis, etc., etc., let us know, we will do our best to help you solve your problem, always the best if more people take a look or anything else. If you can imagine something we can help you with, we are experienced developer mostly in Python but also in other languages and we know a lot about packaging stuff for Federa, so anything else you can imagine we can help you with. The main contacts are mailing list for machine learning, IRC channel also for machine learning Federa and I was inspired by Telegram group for Neuro Federa, so Telegram is coming soon in summer as well. Maybe also we will bridge from our city to Telegram. But yeah, feel free to use anything to contact me. We just created a sick group, sorry, fast group, so now we are able to commentate packages in the right way. We are almost ready but thank you very much that I can use a part of your talk for this because I am really new in machine learning sick and as I said I am trying to revive it so I missed the CFP deadline for a vlog so I cannot do my own talk, so I use this two minutes free at all. Thank you very much. Thank you and yeah, I wanted to mention with just, well, did anybody have any questions that didn't get addressed? Yeah, please. I guess one of the major pitfalls that the project has right now. Good, well at least in my opinion in the neuroimaging we are missing some real staple packages. If I was going to do a course which we have done previously as part of human brain mapping there are a couple of pipelines that are going to be the go-tos in terms of I have done an fMRI study. This is how everybody else does it. And what is really great at least in fMRI is that there is lots of open data. So I wanted to, somewhat related to this, I just wanted to push open neuro as a great resource in terms of, you can download and play with real human fMRI data today and they will give you some of the resources. Well, anybody else want to speak to what are some of the limitations or holes in that? Yeah, I think I have the same feeling that if you try to do some real signs that you will find that well, 100 packages are there but I don't know, we talked about phoenix missing, we talked about the other context, there is a cytoscape thing for doing graph plotting and many, many other stuff that is missing. And I think that the number of scientific users of Fedora is not high enough to attain good feedback loop where people complain when stuff is broken and as a result there is stuff missing and the quality sometimes of the stuff in this package is not high enough. And if we make it possible to use Fedora as a providing for experience, I think that this will improve and then there will be a positive feedback loop. That's something that I've certainly experienced is that meeting a lot of people who are using Fedora but they haven't been contributing back, they just didn't know that there was a group that they could potentially be getting some of their contributions passed on. But also in talking, it was great to see Igor's talk about cross distro collaborations. Sometimes we could be working more with neurodebian in terms of if we've got some issues with upstream choices, we could be working quite well with them in terms of getting more people, showing that there's more people that would be interested in getting those changes. So yeah, there's a lot of great, there's a lot of opportunities for getting involved and helping. But yeah, whether we've got complete pipelines for our core focus is questionable. One good thing that is happening is that Python 3 supporting Fedora is really good and Python is the language of science. So if you want to develop software for scientific purposes, then it makes a lot of sense to switch to Fedora and I think we should be able to take advantage of that. Just a comment. Yeah, I wanted to say that there are many, many fields in neuroscience that people talk about, like people analyzing modern behavior and there are so many packages. So also probably reaching out to scientists and asking them what kind of software they're using, what kind of software they would like to have in the distribution, what would make them change their operating system to do something else. Exactly, this is the same for machine learning. I wanted to also just make a quick pitch or at least tell people about BrainHack. So we're kind of focused on packaging and getting things into Fedora, but if you want to learn more and have a community where other people are doing kind of citizen science, BrainHack is a great resource at least for the neuroimaging. They can point you to the open data repositories that exist. In computational neuroscience, its strength is that you're usually doing simulations. So we can generate data, I mean generating data is what we're trying to do with these tools. And I'm not sure some of those, if there's an equivalent to BrainHack for those. In terms of the people building their own EEG equipment at least, Neurotech X is another great site, another great organization. They've got chapters worldwide based in Montreal, but chapters all over. And EEG equipment and at least cheap transcranial direct current stimulation equipment is very easy to build. And there's a lot of good resources for people working on their own. Anything else you want to cover? Yeah, I think one of the problems that we are currently facing is mostly upstream dependency. So on single, leasingly-mated projects only, we're not doing that first. That is the biggest challenge we are facing right now. A lot of good software is pending for packaging due to upstream. As well as a lot of stuff is packaged but not documented. And it's hard to test when you're... Any other questions? I see a hackathon slide. Yeah. I saw that same slide. When is the hackathon? The plan is to package stuff, basically to work on issues, impact software and package more stuff and to have a quick turnaround where you can submit a package and get a review and submit it to body and get five plus ones within five minutes. And then to be... But if people want to get involved and you're not sure where your talent is, 10 a.m. We will also be working at 10 p.m. No, no, no. Please, please join us. And the other thing is, you know, please come talk to me if you're not sure how your skills can be best used. Like there's always, you know, depending on what you're familiar with already, if you want to put those skills to use, I'm telling you there are opportunities. If there's new things that you want to learn. I'd be very happy to point you to... Usually, if you don't have a telegram group, make the list for the product, if you have the role of package, come, if you don't have the role, come. You work with your first package, the product, you're building a sponsor, a package sponsor, and a package sponsor. Yeah, a lot of people in this conference are special package managers. And hey, let's go out and help to the neuroscience. Yeah, yeah. Alright. Thank you so much. Go to the bar. Thank you.