11:03:15 Okay. Hello. I am James, and I guess broadly, I'm interested in just understanding how and why and when neurons do what they do.
11:03:28 I'm a little bit agnostic to whether it's in a brain or in a machine.
11:03:30 But just want to know why they are doing what they are doing.
11:03:34 And so on the brain side, I can sort of split my work up into 3 sections of sort of theory models.
11:03:40 And also, I guess data. I don't do any data.
11:03:42 I don't do any experiments myself. I do analyse data and design experiments.
11:03:48 So in theory, I've my work is about trying to understand what are the constraints a brain might impose upon the representation it chooses to learn along with all the task is to understand how they interact together, and to understand what the neural representation would look like and the modeling side
11:04:05 I'm particularly interested in understanding sequence, generalization. So how do I learn some statistics or a sequence, and apply it to new situations and leading to predictions in Hippocampus interactive cortex? And now I'm working on frontal cortex?
11:04:18 As well. So I spoke about the data side and in terms of understanding machine representations.
11:04:23 I'm interested in the sort of relationship between how these modern machine architectures, are, what they are doing, and also the architectures in the brain and during parallels between the 2, and seeing that can tell us anything about what the brain's doing what machines are doing and lastly my
11:04:38 thank you. My particular interest are heading in the is into Interpret.
11:04:42 AI, can we, using the tools that we have in neuroscience to understand machines better or make machines perform just the same, but with a little bit more.
11:04:52 Ok, Ok, fine on the social side. Yeah. Running, swimming.
11:04:57 Anything outdoors. I can play ping bong as well, and tennis.
11:05:04 Just a couple.
11:05:15 Okay. Hi, I'm taraqk. I'm based at Ucl, and we're interested in brain plasticity which we established yesterday is mechanisms of plasticity for healthy aging, and we have too, lines of work so a lot of my work in the past has been in
11:05:30 rodents, and we look at snappic cellular and systems, mechanisms of plasticity during aging.
11:05:36 And here we're really looking at potential targets where we could be looking at things that are both facilitating, healthy, aging.
11:05:42 And then also aging related diseases. And we do repeated imaging in behaving animals, and we can look on the left.
11:05:49 Here we have cells on the right. We have synapses, and we can image these over time and look and see how things change after we have changes in the environment.
11:05:57 And the second thing that we've been working on more recently is wroteodent models are great until they're not.
11:06:03 And so we're looking at human work to determine real-world risks and potentially develop interventions for healthy aging.
11:06:08 And so here we're using machine learning approaches to look at large-scale population data.
11:06:13 We've been collaborating with the UN. Who has an incredibly large yet often incomplete set of data globally looking at older populations, and our recent focuses have been looking at risks, things like exercise, loneliness, mental health and workforce engagement as people throughout the lifespan
11:06:32 and particularly how that influences aging, and that allows us to test community-based interventions and potentially facilitate healthy aging.
11:06:40 And my interests are food and wine tasting, but then also running and sorting, thanks.
11:06:52 I get another.
11:06:55 Alright!
11:06:58 Alright!
11:07:03 Hello, everybody! My name is Sebastian. I work at Cesa, in the northeast of Italy.
11:07:10 And yeah, I'm more on the statistical learning side of things.
11:07:13 I guess I come from statistical physics these last couple of years.
11:07:16 I've mostly worked on the theory of artificial neural networks, so like on this what neural networks learn from their data?
11:07:22 Okay, so how does the order do the statistics of data shape the representations that you learned the performance that you achieve?
11:07:27 And for a while I've been looking at sort of good Gaussian models of good pairwise models of data and now I'm increasingly looking sort of it on Gaussian fluctuation.
11:07:35 So I think the topic of this workshop is very topical.
11:07:40 And more recently, I've also been working with the neuroscientists at Cesar.
11:07:43 So we've got some collaborations with the experimentalists looking at sort of recurrent models of memory.
11:07:50 And so, yeah, basically, I'm looking forward to learning as much about neuroscience.
11:07:53 Now, in the next 3 weeks, I actually know what I'm talking about in these meetings.
11:07:58 I also like sport. I also like wine I saw there's lots of piano standing around here.
11:08:04 Now I you know I haven't touched the piano longer than that care to admit, but if any of you want to do some music, I don't know.
11:08:08 Sing, or if you have an instrument with you, maybe we can try to work something out.
11:08:21 Yes.
11:08:23 Thank you. Are you set alright? I'm Gasper.
11:08:30 I'm Odyssey, Austria, very broadly interested in information processing, in biological networks, whatever that means.
11:08:37 But it also includes this type of stuff which is about cell signaling regulation, evolution, information theory.
11:08:44 Pancreas is the recent focus. So I, data plus theory is nice.
11:08:49 Now, you can guess about the success rate of me applying for grants with this sort of a broad adjustment.
11:08:54 It doesn't work for very well, but what I wanted to do is I selected 3 recent things that we actually did in neuroscience in the group as highlights.
11:09:04 One of the things I'm very interested in is how to rigorously test optimality, hypothesis, or normative theories.
11:09:11 So there is a statistical framework that we worked out for this, which we actually then also use in other contexts that I enumerated below.
11:09:19 I like thinking of how to push efficient coding beyond sensory periphery, how to include top-down influences, how to make it very dynamic, so that we are not talking just about filter banks and optimizing filters and fit forward filters, so it's also one
11:09:36 of the recent things that we looked at. And then this was a very different flavor, a large scalescale brain dynamics, a very, very simple, non-equilibrium physics model ising model with the feedback loop that shows that you can have neural avalanches
11:09:54 coexisting with oscillations in a very nice time all, and then this can nicely connected to data, even though it's as simple as they show, and in terms of social stuff.
11:10:03 Hiking, swimming wine, tasting and drinking, cycling. That's great!
11:10:09 And the Kid is joining us on July first. Thank you.
11:10:20 Hello! So I'm Elia. I'm an Emery in Atlanta, in physics, in biology department.
11:10:25 There. My group is probably as broad as cash per say, as we're also doing everything from bacteria to diabetes, to how machine learning works and all things like that.
11:10:36 About half of the group works on neuroscience questions as sort of 3 main questions that we are interested in looking at neural codes.
11:10:43 They are becoming very large. Population codes. What is surprising in those things?
11:10:46 We know web server Rs correlations. Should be jumping with joy or should we be just any precisely what we expected?
11:10:53 And so we are trying to understand what is a structure of multivariate population codes.
11:10:59 What are the various universal teachers in those? And so on?
11:11:02 We're doing non-gouian Bayesian learning in animals, specifically, in birds.
11:11:08 Now we're going to humans and to mice as well.
11:11:12 The question there is, can we model how this animal learns all the way from childhood when they acquire sensory motor behaviors?
11:11:20 And then how do they shape them as adults? And there we discovered a couple of interesting things.
11:11:25 We showed, for example, that reinforcement learning cannot explain what the birds do, and there is some Bayesian approaches that actually explain what the birds do.
11:11:35 And the final thing that is worse note in here is we also started working not that long ago on associative learning.
11:11:44 Our collaborator figured out how to do associative learning and warms, and we have showed that they have the same phenomenology, of a similar opinion to larger animals, except that worms are easier, and we see that the standard models don't work there.
11:12:00 And we propose something that does so that I'm happy to talk about this with others, and what I do as a road cycling, hiking, snorkeling, anything. And it's an outdoor is fine.
11:12:12 If you want to please invite me. If you want to do any of that.
11:12:21 Press, one.
11:12:29 Hi! I'm aviel! I'm a postdoc in Murav's lab.
11:12:35 My background is in physics and musics. But now, during my postdoc, I'm studying the mechanisms which underlie several Dance in the auditory, and visual modalities by 0 dependence, I mean the effect of stimulus history on perception and and the main questions i'm
11:12:53 studying during my post. Doc, are and when their dependence occurs, which levels of processing are involved, low level, high levellevel contributions.
11:13:06 And how do they interact from the bottom up and top-down contributions?
11:13:11 Other questions are, what are the dynamics? The temporal structure of serial dependence?
11:13:18 And if there is a shared mechanism across modalities and across stimulus complexity, and then maybe use a psychophysical experiments on humans for studying those questions, thank you.
11:13:40 Hi, everyone! I'm Kishore. We are very interested in the fact that we think that animals are a lot smarter than we think right.
11:14:03 And so we're fascinated about the ability to potentially reveal the fact that animals exhibit insightful moments, and they have strategies during the learning process.
11:14:13 And so with humans, right, we can think about what we say, what we do and what we know and what we know can often be residing below the surface and hard to kind of access right?
11:14:21 And so we think about this on a spectrum of explicit Tolayton knowledge.
11:14:25 And then as an experimenter, we can try to ask people, and maybe trust what they say or observe their actions right.
11:14:32 And that's essentially where we're stuck in animal neuroscience, right? Observing and interpreting those actions.
11:14:37 But what we want to know is sort of what's residing below the surface.
11:14:39 Right? Is there any evidence of something more interesting going on? And so to do that we sort of do the types of things that are now kind of common in systems, neuroscience?
11:14:48 We try to design behaviors that let us probe animals task knowledge, right?
11:14:53 We have some, I think, clever tricks to do that.
11:14:56 We do that in both single and multitask learning and head fixed and freely moving, we started to design what you see here, which is sort of a volitional homecage environment, where animals can learn up to 1015 tasks.
11:15:05 And then the key from a neural circuit in dynamics perspective is that we think we need to watch the network over time cortical networks, the same exact neurons neuromodulatory axons that are projecting to these cortical networks and then use dimensionality reduction to try
11:15:19 to understand what the heck is this mess telling us now, this last piece is something that I'm really excited to be here for, which is we love to collaborate with theorists right?
11:15:27 Because while we'd love to do that in our own lab, and we can't try to do all of this.
11:15:30 And so, if anybody's interested in these types of questions, please do reach out and from a fun perspective, love to play ping-pong.
11:15:38 I like to eat and drink, and of course chat. So you know, hopefully we can do all that. Well, here.
11:15:52 Oh, good!
11:15:57 It's okay. Good. So I'm an alchiniedman.
11:16:02 I'm from the Whiteiteman Institute, and I care about the design and structure of networks, of neurons and networks, of brains and other biological networks.
11:16:10 The first thing I'll mention is kind of trying to look at the structure and the nature of neural codes where we build.
11:16:17 So statistical models of large population trying to figure out how they are built, what drives them?
11:16:22 Can we learn metrics to infer what they mean for things we haven't seen, and if we can actually try and even synthesize new ones that will carry stuff that we don't know about, we're also interested in the hardware, so we're trying to make links between the structure of
11:16:37 neural networks, and what they actually do, which means we're trying to identify architectural feature that will predict what particular networks will be able to do.
11:16:45 We do that for a lot of simulated networks, but also have been playing a lot lately with real connections where we actually try and see if we can generate connections that are similar to real ones, and also ask if what they would actually perform on how they do things compared to role connectors maybe
11:17:03 even try and manipulate them. And the third thing and the core movies are not working is that we're interested in collective behavior and collective learning in animals.
11:17:13 So we looked fish and mice and people, and also a lot of artificial agents trying to identify the principles that govern how they behave and how they learn.
11:17:24 And my wife and 2 kids will join in 3 weeks, and as you've seen, I like to kill Bonsai. Thank you.
11:17:35 Alright, so what we will do now is that the next thing in the program is the the speed dating.