1 00:00:36,280 --> 00:00:37,400 Well, Gustavo, I've been 2 00:00:37,400 --> 00:00:41,680 reading his work since I was an undergraduate. 3 00:00:41,680 --> 00:00:42,720 I think of my Ph.D. 4 00:00:42,720 --> 00:00:47,760 student and a great expert in network connectivity in the brain 5 00:00:47,760 --> 00:00:50,840 where he was to apply this knowledge of neuroscience 6 00:00:50,840 --> 00:00:53,960 and physics together. 7 00:00:53,960 --> 00:00:56,080 So he's currently full professor 8 00:00:56,080 --> 00:01:00,000 catedrático at the University of Pompeu Fabra in Barcelona, where he's based, 9 00:01:00,000 --> 00:01:03,680 and he also runs the Center for Brain Cognition there 10 00:01:03,760 --> 00:01:08,360 and is an accredited professor and has had an EOC. 11 00:01:08,360 --> 00:01:12,360 Advance Grant is now part of the new Synergy 12 00:01:12,360 --> 00:01:17,640 Grant in neuroscience to be awarded in Spain and to work on 13 00:01:17,720 --> 00:01:22,960 basically stroke recovery in networks and in stroke recovery. 14 00:01:23,040 --> 00:01:27,760 And I was in Hamburg yesterday, and I just found out that Gustavo 15 00:01:27,840 --> 00:01:30,920 is a member of the Science Academy of Hamburg. 16 00:01:31,040 --> 00:01:33,280 So the list is very long. 17 00:01:33,280 --> 00:01:36,560 There also worked with Siemens and has an award from Siemens in terms 18 00:01:36,560 --> 00:01:39,560 of innovation for patents as well when we were talking about this idea. 19 00:01:39,600 --> 00:01:40,480 So yeah, 20 00:01:40,480 --> 00:01:43,720 and for those of you that know 21 00:01:43,920 --> 00:01:48,560 e-brains and so the Human Brain Project legacy and Gustavo was part 22 00:01:48,560 --> 00:01:53,160 of the Human Brain project is now a member of the scientific committee 23 00:01:53,240 --> 00:01:56,360 of e-brains, which there are two people per country. 24 00:01:56,360 --> 00:02:01,640 So I think it's pretty fair to say that, you know, he's not the not maybe. 25 00:02:01,800 --> 00:02:02,760 Well, we don't know. 26 00:02:02,760 --> 00:02:05,560 I would say the top neuroscientist in Spain at the moment. 27 00:02:05,560 --> 00:02:06,880 Yeah. 28 00:02:06,880 --> 00:02:08,640 So we're very happy to have him. 29 00:02:08,640 --> 00:02:16,680 So over to you. Thanks. 30 00:02:16,760 --> 00:02:16,960 Yeah. 31 00:02:16,960 --> 00:02:22,240 Thank you so much for the exaggerated and kind introduction, Brian. 32 00:02:22,280 --> 00:02:26,600 And I think especially for for inviting me and for having the possibility 33 00:02:26,600 --> 00:02:31,080 to present part of my interest nowadays 34 00:02:31,160 --> 00:02:34,040 from the scientific point of view. 35 00:02:34,040 --> 00:02:37,320 So the title of my talk is a little bit 36 00:02:37,320 --> 00:02:41,040 provocative as you realize the thermodynamics of mine. 37 00:02:41,040 --> 00:02:47,640 I mean, I hope that I think I have never found a paper 38 00:02:47,640 --> 00:02:51,440 on thermodynamics and neuroscience before. 39 00:02:51,520 --> 00:02:56,120 So from that point of view, it seems to be really a little bit provocative. 40 00:02:56,120 --> 00:02:58,560 But at the end of the day, you will see that the content 41 00:02:58,560 --> 00:02:59,960 is not provocative at all. 42 00:02:59,960 --> 00:03:03,440 As I mean, it's a very, very, very robust 43 00:03:03,680 --> 00:03:07,960 framework, very natural and very intuitive framework that 44 00:03:07,960 --> 00:03:11,240 we think can be used in neuroscience 45 00:03:11,240 --> 00:03:14,840 for answering some particular questions. 46 00:03:14,920 --> 00:03:17,920 And this is how I will start 47 00:03:18,040 --> 00:03:20,800 it just motivating the questions. 48 00:03:20,800 --> 00:03:25,280 So what we want to achieve, which kind of question we want to answer 49 00:03:25,360 --> 00:03:28,760 in this cognitive neuroscience field, 50 00:03:28,840 --> 00:03:31,680 the main question that we are 51 00:03:31,680 --> 00:03:35,600 trying to solve is a about 52 00:03:35,680 --> 00:03:36,840 the brain 53 00:03:36,840 --> 00:03:40,800 dynamics and who is running the brain dynamics? 54 00:03:40,880 --> 00:03:43,040 Who is orchestrating the brain dynamics? 55 00:03:43,040 --> 00:03:46,320 And this is what we call more technically the data 56 00:03:46,400 --> 00:03:49,000 organization in the brain dynamics. 57 00:03:49,000 --> 00:03:52,400 So just to give you a feeling, we want to know very, 58 00:03:52,400 --> 00:03:53,600 very fundamental things. 59 00:03:53,600 --> 00:03:58,320 For example, if I am doing nothing, what we call our data resting state, 60 00:03:58,400 --> 00:04:03,440 how is my hierarchical organization at the spatial temporal level? 61 00:04:03,440 --> 00:04:07,560 So they are all regions 62 00:04:07,640 --> 00:04:08,840 equally important. 63 00:04:08,840 --> 00:04:12,720 So we have a kind of democracy, we have a flat organization 64 00:04:12,800 --> 00:04:16,600 or there is a kind of in the other extreme dictatorship, there are some regions 65 00:04:16,600 --> 00:04:22,160 we are really running the shows and giving and driving. 66 00:04:22,160 --> 00:04:26,280 The rest of the brain is one question And then how would this theoretical 67 00:04:26,280 --> 00:04:30,600 organization change and differentiate to other, for example, cognition? 68 00:04:30,600 --> 00:04:34,960 So instead of doing nothing, I start to do a working memory task 69 00:04:35,000 --> 00:04:38,000 or decision making social task, whatever. 70 00:04:38,240 --> 00:04:40,640 So is something changing or not? 71 00:04:40,640 --> 00:04:44,240 I mean, from the hierarchical point of view and of course in different 72 00:04:44,240 --> 00:04:50,160 brain states, if I am sleeping, if I am under anesthesia, if I take drugs 73 00:04:50,240 --> 00:04:53,240 or in the biomedical context, 74 00:04:53,360 --> 00:04:56,920 if I am healthy, or I have 75 00:04:57,000 --> 00:04:58,880 some kind of neuropsychiatric 76 00:04:58,880 --> 00:05:03,040 disease, is this Iraqi colonization reflecting that? 77 00:05:03,120 --> 00:05:07,920 And if it is reflecting that is perhaps the causes of my disease or not? 78 00:05:08,000 --> 00:05:12,080 And if I know how, how is the right hierarchical organization 79 00:05:12,080 --> 00:05:14,560 and how is the brain, the irregular indecision 80 00:05:14,560 --> 00:05:20,680 in a particular disease perhaps, that give me some hints to not only predict 81 00:05:20,680 --> 00:05:25,880 and and and make some diagnosis on that, but even to design 82 00:05:25,880 --> 00:05:29,960 some possible therapies. 83 00:05:30,040 --> 00:05:32,720 So the the most natural way 84 00:05:32,720 --> 00:05:36,800 of starting to talk about the hierarchy actually is nothing new. 85 00:05:36,800 --> 00:05:37,480 I mean, people 86 00:05:37,480 --> 00:05:42,280 in artificial intelligence, in computer science, they know that it seems 87 00:05:42,360 --> 00:05:45,840 well, not centuries, but decades at least. 88 00:05:45,920 --> 00:05:46,640 And is it? 89 00:05:46,640 --> 00:05:50,440 Well, then just do it and measure your functionality and 90 00:05:50,480 --> 00:05:53,120 direction your interactions 91 00:05:53,200 --> 00:05:54,640 because you are interesting dynamics. 92 00:05:54,640 --> 00:05:57,600 So it should be a special temporal interactions. 93 00:05:57,600 --> 00:06:00,760 Just measure with your favorite measure 94 00:06:00,800 --> 00:06:04,840 of function and interactions and then put that in a graph and do it. 95 00:06:04,840 --> 00:06:06,920 You don't kill analysis, nothing else. 96 00:06:06,920 --> 00:06:09,320 It's nothing to to complex. 97 00:06:09,320 --> 00:06:11,360 I mean, all the tools are there. 98 00:06:11,360 --> 00:06:13,880 The problem is that's the type of data that we have 99 00:06:13,880 --> 00:06:17,520 in neuroscience are not always allowing this thesis. 100 00:06:17,520 --> 00:06:20,160 Try forward, detect the solution 101 00:06:20,160 --> 00:06:22,800 sometime is possible and I will show in the next slide. 102 00:06:22,800 --> 00:06:26,800 One example that is this is what they call the the detect 103 00:06:26,880 --> 00:06:29,880 this the most simple solution. 104 00:06:29,920 --> 00:06:33,520 For example, you can use Granger causality. 105 00:06:33,520 --> 00:06:39,200 We were talking about that at a couple of minutes ago and 106 00:06:39,280 --> 00:06:42,960 undefined your functional interaction in term of Granger causality 107 00:06:42,960 --> 00:06:47,760 and then you build your graphs under all the different situation 108 00:06:47,760 --> 00:06:48,760 and they start to analyze. 109 00:06:48,760 --> 00:06:54,240 Look, I mean in this conditions, that particular region is really 110 00:06:54,320 --> 00:06:56,120 running the culture. 111 00:06:56,120 --> 00:06:59,040 So what the whole talk is about, 112 00:06:59,040 --> 00:07:04,760 what we can do if we don't want to measure directly the functional interaction 113 00:07:04,760 --> 00:07:09,640 because we don't have enough data, that is the main problem. 114 00:07:09,720 --> 00:07:13,520 And we are using the three coming from the thermodynamics, 115 00:07:13,600 --> 00:07:17,960 which is sounds very technical, but you will see it's very, very simple. 116 00:07:17,960 --> 00:07:20,320 It's the breaking of the detail balance. 117 00:07:20,320 --> 00:07:21,880 So we know in the thermodynamic. 118 00:07:21,880 --> 00:07:25,640 But I will describe that with some details that in the talk 119 00:07:25,720 --> 00:07:30,080 that when you break the symmetry in the interactions, 120 00:07:30,160 --> 00:07:32,880 meaning if you break the symmetries, you start 121 00:07:32,880 --> 00:07:35,880 to build up hierarchical organization, 122 00:07:35,920 --> 00:07:39,760 then this is what we call the breaking of the detail balance. 123 00:07:39,920 --> 00:07:43,200 And that has consequences 124 00:07:43,280 --> 00:07:45,880 that could be reflected in other measurements 125 00:07:45,880 --> 00:07:47,520 which are not the functional interaction. 126 00:07:47,520 --> 00:07:52,280 For example, one very well-known measure is the arrow of time, 127 00:07:52,480 --> 00:07:54,560 and that is what they will use 128 00:07:54,560 --> 00:07:58,440 just by concentrated on the arrow of time of brain signals. 129 00:07:58,520 --> 00:08:01,680 Indirectly, I can say something 130 00:08:01,680 --> 00:08:05,360 about the hierarchy organization and that is fantastic 131 00:08:05,440 --> 00:08:09,080 because as we will see, the collective decision of that of time 132 00:08:09,160 --> 00:08:13,600 is a much more simple business to measure with the 133 00:08:13,680 --> 00:08:18,840 with a low amount, a small amount of data. 134 00:08:18,920 --> 00:08:20,560 And that is the talk. 135 00:08:20,560 --> 00:08:25,120 I mean, basically to to see how we can measure the theoretical 136 00:08:25,160 --> 00:08:31,200 or any session indirectly through this feature of the done while in time, namely 137 00:08:31,280 --> 00:08:36,640 that what we call out of time or non reversibility or non-equilibrium. 138 00:08:36,800 --> 00:08:41,840 So we we practically equalize all these words out of time, 139 00:08:41,840 --> 00:08:42,640 not of activity. 140 00:08:42,640 --> 00:08:46,920 The non-equilibrium and what is important for us is, you know, to get organization. 141 00:08:46,920 --> 00:08:50,720 I don't care if the brain is in article by like because I physicist but 142 00:08:50,800 --> 00:08:54,600 but I theoretical immunoassay I don't care if that is a narrow of time 143 00:08:54,600 --> 00:08:56,400 or it's just Barcelona with it 144 00:08:56,400 --> 00:09:00,040 but they care about the target organization and that is what we do. 145 00:09:00,120 --> 00:09:04,120 So in the first part that I, I will do this in a model three way 146 00:09:04,120 --> 00:09:07,760 and then at the end of the talk, I will say a couple of what about that 147 00:09:07,840 --> 00:09:11,880 we can use now these new measurements instead of the classical measure 148 00:09:11,880 --> 00:09:15,760 like function are going activity to job correlations 149 00:09:15,840 --> 00:09:16,200 or even 150 00:09:16,200 --> 00:09:20,040 sophisticated measure in cases that we can use that 151 00:09:20,120 --> 00:09:24,400 as I mentioned, greater causality before to construct a model. 152 00:09:24,480 --> 00:09:29,280 And then we will have some causal mechanistic explanation 153 00:09:29,360 --> 00:09:33,600 of this particular type of hierarchical organization. 154 00:09:33,680 --> 00:09:35,120 So that is the motivation. 155 00:09:35,120 --> 00:09:38,720 I'm I'm glad of that though, first very rapidly, 156 00:09:38,720 --> 00:09:42,640 because if you are in neuroscience, I mean, people were interested 157 00:09:42,640 --> 00:09:46,680 in the local organization, the whole brain level 158 00:09:46,760 --> 00:09:48,800 also since many decades. 159 00:09:48,800 --> 00:09:53,280 One typical example that we know is the standard. 160 00:09:53,280 --> 00:09:57,080 Can that together with the champion 161 00:09:57,080 --> 00:10:01,000 sans your use, it is this concept from bars 162 00:10:01,080 --> 00:10:06,120 that was a global workspace idea and it's very well reflected in this cartoon. 163 00:10:06,120 --> 00:10:12,120 I mean where would you see all these nodes, which are the brain regions 164 00:10:12,200 --> 00:10:14,600 and and the imagine it 165 00:10:14,600 --> 00:10:19,240 still strange regional hierarchy you know organize it where you have very physical 166 00:10:19,240 --> 00:10:23,800 brain regions probably more associated with sensory information processing 167 00:10:23,880 --> 00:10:27,160 and some central regions which are this global workspace 168 00:10:27,160 --> 00:10:32,080 which they are in the central circle which are regular 18 hour 169 00:10:32,120 --> 00:10:37,200 driving at orchestrate, orchestrating the the other regions. 170 00:10:37,280 --> 00:10:41,160 And in fact, as you know, they use this theory as an example. 171 00:10:41,280 --> 00:10:45,000 It's not necessarily the the main use of that 172 00:10:45,080 --> 00:10:49,200 for plane in a different brain instead like consciousness. 173 00:10:49,280 --> 00:10:54,800 So if the global workspace those of you the region has the top of the Iraqi 174 00:10:54,800 --> 00:10:59,000 allow the ignition of information so the transfer of information 175 00:10:59,000 --> 00:11:02,160 from the periphery to another part of the brain to another 176 00:11:02,160 --> 00:11:06,240 part the in the in the periphery than 177 00:11:06,320 --> 00:11:09,320 they claim, then there are some evidences 178 00:11:09,360 --> 00:11:12,960 that is associated with what we call consciousness. 179 00:11:12,960 --> 00:11:16,880 And if you are not allowing this this ignition that you see, 180 00:11:16,880 --> 00:11:22,120 they are in this read the example called 60 on the right 181 00:11:22,200 --> 00:11:25,440 called subliminal that you see are some activation 182 00:11:25,440 --> 00:11:28,600 but is not enough to 183 00:11:28,680 --> 00:11:31,560 to to to a start 184 00:11:31,560 --> 00:11:35,920 to this huge integration that you see for example in the green curve. 185 00:11:36,000 --> 00:11:39,560 Okay as I said, the most simple and direct 186 00:11:39,560 --> 00:11:43,360 way is really to measure functional interaction. 187 00:11:43,440 --> 00:11:47,000 We took an example where we were able to do that 188 00:11:47,080 --> 00:11:51,560 because we had a lot of data and that is the case of the Human 189 00:11:51,560 --> 00:11:57,120 Connection Project, as you know, is a large collection of a thousand people, 190 00:11:57,200 --> 00:12:00,920 a thousand participants, many different condition. 191 00:12:00,920 --> 00:12:04,200 Neuroimaging is if I mainly the result is mean, 192 00:12:04,400 --> 00:12:08,760 but I will concentrate only on them if MRI today 193 00:12:08,840 --> 00:12:10,040 and there are many conditions, 194 00:12:10,040 --> 00:12:13,800 resident state and many different cognitive condition that you see here. 195 00:12:13,800 --> 00:12:18,160 The Tauber like working memory, socio relational task model task 196 00:12:18,200 --> 00:12:21,200 language gambling. Imagine at this. 197 00:12:21,400 --> 00:12:24,800 And then if you take your measure of Granger causality, 198 00:12:24,800 --> 00:12:29,600 I took a transfer entropy normalize it with silvergate blah blah blah. 199 00:12:29,600 --> 00:12:33,080 I don't go into the details. 200 00:12:33,160 --> 00:12:36,840 Then you can measure directly the interactions 201 00:12:36,920 --> 00:12:39,720 and then you select who are at the top of the hierarchy. 202 00:12:39,720 --> 00:12:43,880 And those are the red regions in the top renderings. 203 00:12:43,960 --> 00:12:48,640 In each different conditions, you see that there are different regions 204 00:12:48,720 --> 00:12:52,760 running the Joe They are the I don't remember, I think I took a 205 00:12:52,760 --> 00:12:56,680 they are like five or 10% of the top regions 206 00:12:56,760 --> 00:12:58,840 and that is and do have more if you do. 207 00:12:58,840 --> 00:13:03,920 Now the intersection is a little bit more sophisticated at the intersection, 208 00:13:04,000 --> 00:13:05,040 but there is 209 00:13:05,040 --> 00:13:08,880 just roughly to give you a flavor is and it kind of intersection 210 00:13:08,880 --> 00:13:12,960 of all these red regions all the top regions in 211 00:13:12,960 --> 00:13:16,840 in different cognitive situations than what you get is 212 00:13:16,880 --> 00:13:22,240 what you see in B in this rendering this yellow and orange regions 213 00:13:22,320 --> 00:13:27,160 which are basically quantification of the global workers space. 214 00:13:27,240 --> 00:13:30,640 What the regions with and all possible 215 00:13:30,640 --> 00:13:35,160 cognitive demands are always running the shows 216 00:13:35,240 --> 00:13:37,880 now and the folk would be associated with that. 217 00:13:37,880 --> 00:13:41,400 So from that point of view, in that case, because we have so many 218 00:13:41,400 --> 00:13:45,440 and so high quality data, remember not only 219 00:13:45,440 --> 00:13:49,520 the large number of participants, but also two sessions, in some cases 220 00:13:49,520 --> 00:13:54,560 they are four session of resting state, long sessions, 22 minutes, 221 00:13:54,640 --> 00:13:58,520 and with the fabulous year of 0.7 seconds. 222 00:13:58,520 --> 00:14:00,400 So it's fantastic data. 223 00:14:00,400 --> 00:14:03,200 So with that you are 224 00:14:03,280 --> 00:14:07,440 you managed to really calculate the convention causality. 225 00:14:07,520 --> 00:14:11,000 Go now to another context which is also equally relevant 226 00:14:11,200 --> 00:14:14,200 psychiatric application, biomedical obligation. 227 00:14:14,280 --> 00:14:18,600 You are lucky if you get the resting state of 7 minutes 228 00:14:18,680 --> 00:14:22,200 with the bacteria from the hospital, 2 seconds 229 00:14:22,200 --> 00:14:25,920 and 20 subjects. 230 00:14:26,000 --> 00:14:28,400 You cannot calculate anything down. 231 00:14:28,400 --> 00:14:33,440 So therefore we were motivated and I want to apply all these ideas 232 00:14:33,440 --> 00:14:34,720 to to psychiatry. 233 00:14:34,720 --> 00:14:39,600 So I am obliged to develop a framework, a theoretical framework 234 00:14:39,600 --> 00:14:43,760 that allows me with this low amount of data to characterize 235 00:14:43,840 --> 00:14:46,280 and this is what we call that the more dynamic state of mind 236 00:14:46,280 --> 00:14:50,040 and now I will go a little bit more into the details. 237 00:14:50,120 --> 00:14:55,080 Before that, I suddenly started to have a result. 238 00:14:55,160 --> 00:14:58,240 Originally, I am coming from quantum mechanics, I mean actually 239 00:14:58,240 --> 00:15:03,600 from electrodynamics, quantum mechanics and a I and I was always even nowadays 240 00:15:03,600 --> 00:15:07,920 very interested in history of science and what you see 241 00:15:08,000 --> 00:15:12,680 there is the the, the image of a very well-known 242 00:15:12,760 --> 00:15:17,520 scientist, Edwin Schrodinger, a Nobel Prize in quantum mechanics. 243 00:15:17,600 --> 00:15:22,280 And of course, I mean he he had he ran very beautiful 244 00:15:22,280 --> 00:15:25,720 and very relevant ideas in quantum mechanics. 245 00:15:25,840 --> 00:15:27,200 And I got a question. 246 00:15:27,200 --> 00:15:31,000 For example, I was not aware that he was working also 247 00:15:31,000 --> 00:15:34,000 at the end of his life in in life science. 248 00:15:34,120 --> 00:15:40,240 And he had a great idea that he wrote in this book what this life. 249 00:15:40,320 --> 00:15:44,840 And it's a tiny book, very beautiful, written. 250 00:15:44,840 --> 00:15:49,840 And there he puts an idea, namely that we should use 251 00:15:49,920 --> 00:15:53,480 non-equilibrium thermodynamics in the context of biology 252 00:15:53,480 --> 00:15:57,600 because he associate basically life with non-equilibrium. 253 00:15:57,680 --> 00:16:01,520 But that's open a whole field in system biology. 254 00:16:01,600 --> 00:16:06,480 So at the molecular level where they they started to apply really 255 00:16:06,480 --> 00:16:10,800 those ideas of non-equilibrium thermodynamics in protein 256 00:16:10,800 --> 00:16:13,800 synthesis and so on, and continue even nowadays. 257 00:16:14,000 --> 00:16:19,080 So what I am doing today here is to try to apply exactly 258 00:16:19,080 --> 00:16:23,680 the same kind of ideas but in the macroscopic context. 259 00:16:23,800 --> 00:16:27,160 So I am not applying this to molecules, but I will apply to the whole way 260 00:16:27,240 --> 00:16:30,240 that is the 261 00:16:30,360 --> 00:16:33,600 joy attached to this claim that the idea is not coming from me. 262 00:16:33,600 --> 00:16:37,680 It's coming from him. And 263 00:16:37,760 --> 00:16:38,720 so here is a 264 00:16:38,720 --> 00:16:42,000 cartoon session of this idea now with a little bit more of that there. 265 00:16:42,120 --> 00:16:45,840 What you see on the top is a brain with just a cartoon. 266 00:16:45,840 --> 00:16:46,880 I said four regions. 267 00:16:46,880 --> 00:16:50,600 Of course, we will take hundred of regions most of the time. 268 00:16:50,600 --> 00:16:56,320 And then you if you study the the I introduction, 269 00:16:56,560 --> 00:16:59,400 assuming that we can measure that 270 00:16:59,400 --> 00:17:03,800 and if we know that the functional interaction are absolutely symmetrical, 271 00:17:03,880 --> 00:17:07,200 then automatically we know that the system is nonhierarchical. 272 00:17:07,280 --> 00:17:10,280 This is a democratic flat system. 273 00:17:10,280 --> 00:17:12,320 They have all the same problem. 274 00:17:12,320 --> 00:17:14,920 There is no a leader there. 275 00:17:14,920 --> 00:17:18,360 A From the thermodynamic point of view, this is called 276 00:17:18,440 --> 00:17:22,560 detail balance, this symmetry in the information flow. 277 00:17:22,640 --> 00:17:26,200 And then we know automatically two consequences. 278 00:17:26,200 --> 00:17:31,000 One that's that is no entropy production. 279 00:17:31,080 --> 00:17:32,720 And second, which is actually 280 00:17:32,720 --> 00:17:37,120 a theology of the first, that is an arrow of time. 281 00:17:37,120 --> 00:17:39,680 So that is that there is no out of time. 282 00:17:39,680 --> 00:17:43,200 Sorry, that is reversibility. 283 00:17:43,280 --> 00:17:47,280 If you go now to the bottom part of the slide 284 00:17:47,360 --> 00:17:49,680 and then you break the order, assumed that 285 00:17:49,680 --> 00:17:52,960 you are breaking the symmetry 286 00:17:53,040 --> 00:17:56,760 that you start to generate those rings like in the global workers base. 287 00:17:56,840 --> 00:18:01,800 So some of the central guys are more important because they're 288 00:18:01,880 --> 00:18:04,360 they could be driving the other guys. 289 00:18:04,360 --> 00:18:10,440 And then we know that if we break the symmetry 290 00:18:10,520 --> 00:18:13,520 in the functional interaction we have, yet our key 291 00:18:13,560 --> 00:18:16,320 and we have a breaking of the little violence 292 00:18:16,320 --> 00:18:19,280 and then exactly the control that we assumed before, 293 00:18:19,280 --> 00:18:23,440 we have non-equilibrium and that means that we have production 294 00:18:23,440 --> 00:18:28,520 entropy different from zero and we have non reversibility. 295 00:18:28,520 --> 00:18:31,920 Another of that and that is the good new because 296 00:18:32,040 --> 00:18:35,040 the production entropy is something that is very beautiful. 297 00:18:35,200 --> 00:18:38,200 But we can also not measure per say 298 00:18:38,280 --> 00:18:41,120 that paper is measured in that even from my lab. 299 00:18:41,120 --> 00:18:45,960 But we do dirty tricks that I am ashamed to talk about that today. 300 00:18:46,040 --> 00:18:48,960 We reduce the space, I mean, and using technique. 301 00:18:48,960 --> 00:18:50,520 And so 302 00:18:50,600 --> 00:18:52,160 but these are high dimensional systems. 303 00:18:52,160 --> 00:18:56,400 So to to estimate probability and entropy in the high dimensional system, 304 00:18:56,400 --> 00:18:59,400 forget it. But the idea 305 00:18:59,600 --> 00:19:04,480 but as I say, they were of that is the arrow of time. 306 00:19:04,560 --> 00:19:07,200 And this is what we measure. 307 00:19:07,200 --> 00:19:12,360 We just look, if there is a distinction in the in the time evolution of our 308 00:19:12,360 --> 00:19:15,920 but in signals, most of the time could be bold signals. 309 00:19:16,000 --> 00:19:19,240 I will show some example with the local field but then Jensen 310 00:19:19,240 --> 00:19:24,360 so that you get the feeling that it's not depending on the other measurement. 311 00:19:24,440 --> 00:19:28,480 So this is just the Hollywood version of the concept of not reversibility 312 00:19:28,640 --> 00:19:30,080 and reversibility. 313 00:19:30,080 --> 00:19:32,360 Just to give you a feeling that what you see 314 00:19:32,360 --> 00:19:35,680 in these two movies on the left is the non reversible system. 315 00:19:35,680 --> 00:19:38,680 Non-Equilibrium So a glass of wine 316 00:19:38,680 --> 00:19:42,680 which is destroyed by the impinging bullets 317 00:19:42,760 --> 00:19:46,360 that you see the for one movie and the same movie, but backwards 318 00:19:46,440 --> 00:19:49,280 and it's very clear what is for What about this fog one. 319 00:19:49,280 --> 00:19:50,400 Yeah. 320 00:19:50,400 --> 00:19:54,760 A On the other side you have a typical classical 321 00:19:54,840 --> 00:19:59,320 reversible system colliding consisting of volleyballs 322 00:19:59,400 --> 00:20:02,800 that would be perfectly elastic is perfectly reversible. 323 00:20:02,880 --> 00:20:05,720 In fact, what you see is the moving forward backward. 324 00:20:05,720 --> 00:20:09,520 And you cannot distinguish, you cannot say which one is the reality. 325 00:20:09,560 --> 00:20:12,160 Both that, okay, 326 00:20:12,240 --> 00:20:15,600 the problem is that it's not always the case. 327 00:20:15,680 --> 00:20:19,560 That is so easy to distinguish reversibility from the possibility. 328 00:20:19,760 --> 00:20:22,960 And this is the reason why I put here on the left. 329 00:20:23,040 --> 00:20:27,200 Actually, I have to confess I am a fan of Christopher Nolan, 330 00:20:27,280 --> 00:20:31,360 but Tenet is the worst movie, of course. 331 00:20:31,440 --> 00:20:34,760 But for me it was a good example because are people traveling forward 332 00:20:34,760 --> 00:20:39,440 and backward in time producing entropy or or the increase in entropy? 333 00:20:39,520 --> 00:20:43,160 You see in this excerpt of the movie that there are some people 334 00:20:43,160 --> 00:20:46,480 there, for example, traveling forward and backwards. 335 00:20:46,480 --> 00:20:47,360 It's very clear. 336 00:20:47,360 --> 00:20:51,600 It's like the glass of wine is is absolutely clear. 337 00:20:51,680 --> 00:20:52,880 What you see on the right. 338 00:20:52,880 --> 00:20:55,040 It is a tricky system. 339 00:20:55,040 --> 00:20:57,720 It's a physical system. It's not important in the detail. 340 00:20:57,720 --> 00:21:02,160 It is an expensive too, but that by construction, 341 00:21:02,240 --> 00:21:06,160 just skiing in that nature of physics paper, 342 00:21:06,240 --> 00:21:09,120 it generates a system that he knows 343 00:21:09,120 --> 00:21:12,440 because he simulated is by definition 344 00:21:12,440 --> 00:21:16,440 non reversible, is in non-equilibrium. 345 00:21:16,520 --> 00:21:19,520 And then you measure something, it's the magnetization 346 00:21:19,720 --> 00:21:23,680 and you see how the magnetization evolve in time is one of this kind of actually, 347 00:21:23,880 --> 00:21:28,400 I run the simulation, I run the simulation on the proposed budget game 348 00:21:28,480 --> 00:21:32,840 and I don't remember which one is the forward on the backward ration. 349 00:21:32,840 --> 00:21:36,840 So one caller is for when I'm the other college smuggler, 350 00:21:36,920 --> 00:21:40,880 I don't know and I cannot recognize. 351 00:21:40,960 --> 00:21:43,400 So meaning, even if I know 352 00:21:43,400 --> 00:21:46,680 that I should be able to recognize 353 00:21:46,760 --> 00:21:50,360 is not this is not intuitive A 354 00:21:50,520 --> 00:21:55,160 and the first year since kicking with a fabulous idea 355 00:21:55,160 --> 00:21:58,360 so basically this is assuming all this concept of 356 00:21:58,440 --> 00:22:01,440 out of time entropy in an non-equilibrium or so on. 357 00:22:01,680 --> 00:22:05,480 But the important part of this slide is the the 358 00:22:05,560 --> 00:22:08,560 the panel on the on the right. 359 00:22:08,720 --> 00:22:12,240 What this is reflecting is the in some cases 360 00:22:12,240 --> 00:22:17,720 it's very easy to distinguish like in the top four from backward 361 00:22:17,800 --> 00:22:19,640 A you don't need nothing. 362 00:22:19,640 --> 00:22:25,520 I mean you just need your eyes and that is in other cases like in this 363 00:22:25,600 --> 00:22:30,400 physical example of you are seeing, she is difficult, 364 00:22:30,480 --> 00:22:33,440 but we know how to distinguish things 365 00:22:33,440 --> 00:22:36,440 in difficult cases, especially nowadays. 366 00:22:36,440 --> 00:22:39,720 It is actually the field of 367 00:22:39,800 --> 00:22:42,360 of machine learning. 368 00:22:42,360 --> 00:22:44,520 A And therefore the idea 369 00:22:44,520 --> 00:22:47,920 was to use a machine learning network that we call on purpose. 370 00:22:47,920 --> 00:22:50,920 Then it but means a temporal 371 00:22:50,920 --> 00:22:53,920 evolution deep learning network 372 00:22:54,080 --> 00:22:56,680 in order to distinguish in the case of the brain, 373 00:22:56,680 --> 00:23:01,720 which is not like the case of the movie of Tenet it or the class of one, 374 00:23:01,720 --> 00:23:06,200 but more the difficult case and to use this machine learning technique 375 00:23:06,440 --> 00:23:11,800 in order to just distinguish the direction of the time 376 00:23:11,880 --> 00:23:14,880 to determine if there is a matter of time or not. 377 00:23:15,040 --> 00:23:17,240 So we apply this in all possible contexts. 378 00:23:17,240 --> 00:23:20,480 One can apply this to at the whole brain level, 379 00:23:20,480 --> 00:23:24,120 at the network level, at the local level, 380 00:23:24,200 --> 00:23:27,200 with different type of signals, different type of species. 381 00:23:27,200 --> 00:23:30,320 Of course, at the beginning they just concentrate on ball 382 00:23:30,360 --> 00:23:34,440 signals, neuroimaging in humans and the different conditions. 383 00:23:34,520 --> 00:23:35,240 And what you see in 384 00:23:35,240 --> 00:23:39,240 the middle panel is just in one dimension so that we can plot something 385 00:23:39,320 --> 00:23:44,160 is just your ball signal as you measured it in one region. 386 00:23:44,240 --> 00:23:46,800 A And then by hand 387 00:23:46,800 --> 00:23:51,560 you generate the backward motion, you flip it. Yes. 388 00:23:51,640 --> 00:23:56,160 And then you know which one you flip it and which one is the real measurement. 389 00:23:56,160 --> 00:24:00,880 And then you can train a deep learning classifier black box 390 00:24:00,960 --> 00:24:02,720 in a supervised way 391 00:24:02,720 --> 00:24:06,440 because you know who is or whose role or who is of output. 392 00:24:06,520 --> 00:24:10,680 And then see if you can separate that than using all the tricks 393 00:24:10,760 --> 00:24:14,360 of machine learning in a cross-validation, blah blah, blah blah. 394 00:24:14,440 --> 00:24:18,280 You check the performance if you see that the performance is high. 395 00:24:18,280 --> 00:24:22,960 So you can distinguish a signal that you have never seen before, 396 00:24:23,000 --> 00:24:26,280 that means that that is an arrow of time 397 00:24:26,360 --> 00:24:29,000 that means that the system is not reversible. 398 00:24:29,000 --> 00:24:34,080 That means that this is a non-equilibrium and the most important consequence 399 00:24:34,080 --> 00:24:39,080 is to see them easier to kick out and you have a quantification. 400 00:24:39,080 --> 00:24:44,640 So the performance of this network quantified the level of hierarchy. 401 00:24:44,720 --> 00:24:50,560 On the contrary, if it is low, you have more reversibility, more akin, leave, less 402 00:24:50,640 --> 00:24:52,280 less theoretical organization. 403 00:24:52,280 --> 00:24:56,120 But as I said, you can use that in different context. 404 00:24:56,120 --> 00:24:58,280 I am not going into the details. 405 00:24:58,280 --> 00:25:03,360 What you see on the right as a whole brain level network, labor or local level. 406 00:25:03,360 --> 00:25:06,520 I think I will give you an example of the 407 00:25:06,520 --> 00:25:10,680 of the two extremes global and local level. 408 00:25:10,760 --> 00:25:14,680 So this is one case 409 00:25:14,760 --> 00:25:16,400 again, HDP, 410 00:25:16,400 --> 00:25:20,920 so that thousand people neuroimaging here, but the resting state 411 00:25:20,920 --> 00:25:25,440 and all the different conditions I concentrate on the on the left panel, 412 00:25:25,520 --> 00:25:26,520 it's the important one. 413 00:25:26,520 --> 00:25:31,280 And you see on the x axis, all the conditions rest in first 414 00:25:31,360 --> 00:25:34,040 and then all the cognitive task. 415 00:25:34,040 --> 00:25:38,440 And on the y axis you see for the cross-validation 416 00:25:38,440 --> 00:25:42,880 set, therefore you see more or less hundred dots. 417 00:25:42,960 --> 00:25:44,600 I trained it the ten and 418 00:25:44,600 --> 00:25:47,960 with the more or less 900 participants 419 00:25:47,960 --> 00:25:50,960 and then a 200 420 00:25:51,000 --> 00:25:53,960 random participant for cross-validation. 421 00:25:53,960 --> 00:25:58,920 And you check the the label of non CBT 422 00:25:59,000 --> 00:26:02,520 as a performance of the net for each single subset. 423 00:26:02,600 --> 00:26:05,720 And what you see is somehow expected they are or the reading 424 00:26:05,720 --> 00:26:08,720 increase in order of larger organization 425 00:26:08,760 --> 00:26:13,200 resting is the less dark color, but this hierarchical is not zero. 426 00:26:13,280 --> 00:26:14,800 You see, it's pretty guy. 427 00:26:14,800 --> 00:26:17,200 It's of one five, five, whatever. 428 00:26:17,200 --> 00:26:21,560 So there is a radical realization in your brain, even when you are doing nothing, 429 00:26:21,680 --> 00:26:25,000 which is not cytology, because we knew about the existence of networks. 430 00:26:25,000 --> 00:26:28,760 And so but it's nice to quantify that. 431 00:26:28,760 --> 00:26:33,600 And then what is expected, of course, that when you start to use your brain 432 00:26:33,600 --> 00:26:37,400 for doing something, of course you break even more the symmetry, 433 00:26:37,480 --> 00:26:41,560 even more the functioning director, because you need to promote the cross-talk 434 00:26:41,560 --> 00:26:45,520 very specifically between different members of your brains 435 00:26:45,600 --> 00:26:49,360 and therefore the organization is increasing. 436 00:26:49,400 --> 00:26:52,800 The beauty is that you are quantifying this and then 437 00:26:52,800 --> 00:26:56,360 you can see that there are different target organization 438 00:26:56,440 --> 00:27:00,400 in the social task compared to whatever emotion. 439 00:27:00,400 --> 00:27:04,360 And this is totally different. 440 00:27:04,440 --> 00:27:06,920 I don't comment on the other plots. 441 00:27:06,920 --> 00:27:13,080 This is just to give very rapidly to give you a flavor of biomedical application. 442 00:27:13,160 --> 00:27:15,560 This it is also a public dataset 443 00:27:15,560 --> 00:27:18,560 from UCLA, a 444 00:27:18,640 --> 00:27:20,880 It's very small, but what you have to see on 445 00:27:20,880 --> 00:27:24,320 the top left is exactly the same as we have seen before. 446 00:27:24,320 --> 00:27:29,480 But now for four groups of participants in resting state, 447 00:27:29,720 --> 00:27:35,520 the first group, the Gray is Healthy Group control, and then the other ones 448 00:27:35,520 --> 00:27:40,720 which are significantly below the the control group. 449 00:27:40,800 --> 00:27:43,680 So they have less hierarchical organization 450 00:27:43,680 --> 00:27:47,040 under the same condition reasons they are schizophrenic, 451 00:27:47,040 --> 00:27:50,040 bipolar and attentional deficit. 452 00:27:50,240 --> 00:27:54,080 And then you can go into the more specific description of that. 453 00:27:54,080 --> 00:27:57,080 This is a global measure, of course, of out of time. 454 00:27:57,160 --> 00:28:02,920 If you start to apply exactly the same philosophy at the local brain area level 455 00:28:03,000 --> 00:28:06,080 and then you rendered the differences is what you see. 456 00:28:06,080 --> 00:28:10,560 For example, at the bottom, just the difference between bipolar and controls, 457 00:28:10,560 --> 00:28:14,280 between attentional deficit and controls or schizophrenia and controls. 458 00:28:14,360 --> 00:28:18,440 And then you can start to interpret in which way 459 00:28:18,440 --> 00:28:22,200 that you darker organization change in all of them. 460 00:28:22,200 --> 00:28:26,960 I mean, in all of them globally goes down is flatter but in different ways. 461 00:28:26,960 --> 00:28:30,240 For example, I don't want to go into the details. 462 00:28:30,320 --> 00:28:34,040 You see, schizophrenia in general is going down 463 00:28:34,120 --> 00:28:37,320 all the regions in attentional deficit is mainly 464 00:28:37,320 --> 00:28:41,120 somatosensory or occipital areas going down, which makes sense. 465 00:28:41,120 --> 00:28:44,120 And one can interpret that that because they are more 466 00:28:44,280 --> 00:28:48,320 driven by the vitamin and therefore that then you don't get the gaze 467 00:28:48,320 --> 00:28:50,120 and things like that. 468 00:28:50,200 --> 00:28:51,560 Okay. 469 00:28:51,560 --> 00:28:54,880 Just very rapidly to give you a flavor, 470 00:28:54,880 --> 00:29:01,160 we can simplify this things and instead of using machine learning, 471 00:29:01,240 --> 00:29:04,800 we can use a measure which is even a 472 00:29:04,800 --> 00:29:08,520 much simpler to compute 473 00:29:08,600 --> 00:29:11,600 exactly the same concept out of time. 474 00:29:11,800 --> 00:29:15,120 The idea is following on the top line. 475 00:29:15,120 --> 00:29:21,120 You see how to calculate that for a pair of regions. 476 00:29:21,360 --> 00:29:25,920 And then in the bottom part is you can do this for all pairs 477 00:29:26,000 --> 00:29:29,520 in a matrix and then get the exact information at the whole brain level. 478 00:29:29,640 --> 00:29:31,080 The idea is very simple. 479 00:29:31,080 --> 00:29:33,160 Don't don't see that slide is too complex. 480 00:29:33,160 --> 00:29:35,960 I realize 481 00:29:35,960 --> 00:29:39,960 you measure the correlation but in a shifted way. 482 00:29:40,040 --> 00:29:44,680 So you break the symmetry in time on purpose 483 00:29:44,760 --> 00:29:47,120 of the duration of the two signals. 484 00:29:47,120 --> 00:29:48,160 So instead of calculating 485 00:29:48,160 --> 00:29:51,480 the correlation, as always you do the shifted correlations. 486 00:29:51,560 --> 00:29:54,880 Usually you move one to, I mean, some methodology 487 00:29:54,920 --> 00:30:00,320 for designing the optimal shifting, and then you do this with the forward 488 00:30:00,320 --> 00:30:04,520 version and get the number and then you do the backward maps. 489 00:30:04,520 --> 00:30:07,240 You and do exactly the same. 490 00:30:07,240 --> 00:30:11,120 If the system has an arrow of time, that will be different. 491 00:30:11,200 --> 00:30:15,480 So if you do, the difference of the two shifted correlation, 492 00:30:15,560 --> 00:30:18,000 this is what we call inside out 493 00:30:18,080 --> 00:30:22,320 for some strange reason, but 494 00:30:22,400 --> 00:30:25,400 it's a number and that we characterize the label 495 00:30:25,400 --> 00:30:29,520 of asymmetry in this in this case. 496 00:30:29,600 --> 00:30:32,840 And as I said, you can do that for all possible person 497 00:30:32,840 --> 00:30:35,840 that you have the whole brand label. 498 00:30:35,840 --> 00:30:38,840 Just again, another example in this case, 499 00:30:38,840 --> 00:30:42,800 another species, they are monkeys, another type of measurements. 500 00:30:42,800 --> 00:30:47,400 They are a and 501 00:30:47,480 --> 00:30:51,480 actually intra cortical EEG. 502 00:30:51,560 --> 00:30:56,000 So phase cortical EEG is all high quality 503 00:30:56,200 --> 00:30:59,200 electrical signals for monkeys. 504 00:30:59,360 --> 00:31:03,520 A difference in this case is different brain isolate conditions. 505 00:31:03,600 --> 00:31:05,480 They have manipulation. 506 00:31:05,480 --> 00:31:10,320 Whether you see the difference between awake and asleep or awake 507 00:31:10,320 --> 00:31:15,840 and the action of different anesthetic all under poverty. 508 00:31:15,920 --> 00:31:18,720 And what you see on the right path in all these books, 509 00:31:18,720 --> 00:31:24,440 blood is always the same is that measure of non reversibility meaning 510 00:31:24,520 --> 00:31:26,960 of hierarchy 511 00:31:26,960 --> 00:31:29,880 in all this condition and what you see is in conscious state, 512 00:31:29,880 --> 00:31:33,040 so in awake is always higher than asleep 513 00:31:33,120 --> 00:31:35,520 or higher than in unconscious state, 514 00:31:35,520 --> 00:31:39,400 like in under the effect of anesthetic 515 00:31:39,400 --> 00:31:44,000 or like propofol and D or ketamine and empty. 516 00:31:44,080 --> 00:31:47,520 And then when the effect of the anesthetic on disappear is the third 517 00:31:47,520 --> 00:31:51,880 part of the of the blood that you start to recover 518 00:31:51,880 --> 00:31:55,200 and the organization you start to build up again, 519 00:31:55,280 --> 00:31:59,680 there is only one exception, which is ketamine is going in the other way. 520 00:31:59,760 --> 00:32:01,440 And you can interpret that. 521 00:32:01,440 --> 00:32:04,880 The state reason why ketamine is not only an anesthetic 522 00:32:05,040 --> 00:32:10,120 but is nowadays also use it as a fun drug and 523 00:32:10,200 --> 00:32:13,120 is increasing the hierarchy 524 00:32:13,120 --> 00:32:14,400 in a funny way. 525 00:32:14,400 --> 00:32:17,760 This is a biomedical application of the same framework. 526 00:32:17,760 --> 00:32:20,480 These are data from years and berries. 527 00:32:20,480 --> 00:32:22,640 Coma patients 528 00:32:22,720 --> 00:32:24,600 controls minimal consciousness 529 00:32:24,600 --> 00:32:28,560 and deep coma, and you go in the right direction. 530 00:32:28,560 --> 00:32:30,800 A Of course 531 00:32:30,800 --> 00:32:35,240 controls have a certain level of about you not to give you any session 532 00:32:35,240 --> 00:32:39,320 a little bit less in minimal consciousness and much less significantly, 533 00:32:39,480 --> 00:32:43,480 much less in in the deep learning case 534 00:32:43,560 --> 00:32:46,200 shows 535 00:32:46,200 --> 00:32:49,760 now that we have a good measure that characterized the 536 00:32:49,960 --> 00:32:54,760 the hierarchical organization in 537 00:32:54,840 --> 00:32:58,280 let's see if we can explain that mechanistically. 538 00:32:58,280 --> 00:33:03,040 So what this means with this is a model for the measurement quantification 539 00:33:03,040 --> 00:33:08,200 and we can try now to build a model which explain exactly that measure. 540 00:33:08,200 --> 00:33:13,600 So the the get out of key to the quantification of the Iraqi organization. 541 00:33:13,680 --> 00:33:17,400 And so the 542 00:33:17,480 --> 00:33:21,560 the sense of the model I think is relatively well-known nowadays. 543 00:33:21,560 --> 00:33:25,280 But just to cartoon, is that 544 00:33:25,360 --> 00:33:27,520 what we call whole brain model 545 00:33:27,520 --> 00:33:31,080 and nowadays is mainly the integration 546 00:33:31,080 --> 00:33:35,280 of the coupling coming from the anatomy. 547 00:33:35,360 --> 00:33:39,800 For example, in humans, most of the cases coming through the stratigraphy. 548 00:33:39,800 --> 00:33:43,080 But of course, if you have better ways of describing the anatomy, 549 00:33:43,080 --> 00:33:46,080 you should to use that 550 00:33:46,280 --> 00:33:47,640 is the case of animals. 551 00:33:47,640 --> 00:33:49,520 Most of the cases are structural. 552 00:33:49,520 --> 00:33:53,360 The so-called connectome, the structure are in terms of the fibers, 553 00:33:53,440 --> 00:33:55,520 existing couplings. 554 00:33:55,520 --> 00:33:59,720 And then you put your favorite way of describing the local dynamics 555 00:33:59,960 --> 00:34:03,680 and you try then to explain the global dynamics, nothing else. 556 00:34:03,680 --> 00:34:05,880 This is called hybrid model it, of course. 557 00:34:05,880 --> 00:34:10,840 And you you have to define okay, but which aspect of the global dynamics. 558 00:34:11,000 --> 00:34:14,880 Traditionally, we were always obsessed with the function of connectivity, 559 00:34:14,960 --> 00:34:18,040 functional connectivity dynamics. And so 560 00:34:18,120 --> 00:34:21,120 a what I will use today is 561 00:34:21,280 --> 00:34:25,160 get out again when I say, for example, the inside out matrix. 562 00:34:25,240 --> 00:34:31,600 So the the, the label of of of asymmetry 563 00:34:31,680 --> 00:34:36,040 in the in the forward or backward version of the shifted functional connection. 564 00:34:36,040 --> 00:34:38,840 And I will try to explain that 565 00:34:38,840 --> 00:34:39,920 from the empirical data. 566 00:34:39,920 --> 00:34:43,640 So I will try to fit the model so that they explain the 567 00:34:43,680 --> 00:34:47,960 the so the top matrix is the biblical one and they will 568 00:34:47,960 --> 00:34:52,160 try to generate an explanation of that with the model for that. 569 00:34:52,160 --> 00:34:56,520 They will try in many parameters, namely I will associate to each existing 570 00:34:56,520 --> 00:35:02,120 fiber value and this is what we call effective connectivity. 571 00:35:02,200 --> 00:35:04,840 A And I added just because 572 00:35:04,840 --> 00:35:09,120 the term effective connectivity was used in many different contexts, 573 00:35:09,200 --> 00:35:12,440 I put generative affective connectivity 574 00:35:12,520 --> 00:35:15,680 because it's the one really coming from the from the model. 575 00:35:15,800 --> 00:35:18,000 So and, and we optimized that. 576 00:35:18,000 --> 00:35:20,920 And in this case we use our favorite 577 00:35:20,920 --> 00:35:24,280 the local linear, local nonlinear dynamic, 578 00:35:24,280 --> 00:35:30,640 which is the two Orlando oscillator and does some good reason for using that. 579 00:35:30,720 --> 00:35:34,600 An example first left of again 580 00:35:34,600 --> 00:35:38,640 model three HCB 581 00:35:38,720 --> 00:35:41,960 What you see in green is the box plot of all ties together 582 00:35:42,080 --> 00:35:45,800 now and not distinguishing them but calculated with the inside out. 583 00:35:45,920 --> 00:35:50,000 So they are at the top has the maximal hierarchy 584 00:35:50,080 --> 00:35:54,240 in the middle and blue light blue you see resting state. 585 00:35:54,400 --> 00:35:58,520 The difference that one is with the seven Tesla, the others with a three Tesla. 586 00:35:58,520 --> 00:36:00,160 So but very similar. 587 00:36:00,160 --> 00:36:04,560 And what is important that is they are a below 588 00:36:04,560 --> 00:36:07,840 the hierarchical organization of cognitive tasks 589 00:36:07,880 --> 00:36:12,160 as we found before with with Tenet, but now with the inside out. 590 00:36:12,200 --> 00:36:15,240 So that is is good that meaning that 591 00:36:15,240 --> 00:36:19,600 the methodology on the same data is achieved in exactly the same. 592 00:36:19,680 --> 00:36:22,320 But the most interesting case is the yellow box plot. 593 00:36:22,320 --> 00:36:24,400 The yellow plot is 594 00:36:24,480 --> 00:36:28,800 a condition where they are watching a movie 595 00:36:28,880 --> 00:36:33,760 which is very interesting and use a lot nowadays in neuroscience. 596 00:36:33,840 --> 00:36:37,400 And what we see is something that is unexpected. 597 00:36:37,480 --> 00:36:40,160 The theoretical organization goes down 598 00:36:40,160 --> 00:36:43,920 that caused me a couple of months in my life 599 00:36:44,000 --> 00:36:48,720 because I was absolutely convinced that is an error and I was doing the 600 00:36:48,720 --> 00:36:53,480 MATLAB and so on, say, come on, I am getting older. 601 00:36:53,480 --> 00:36:57,720 I mean, I'm not used to do 602 00:36:57,800 --> 00:36:59,840 I calculated within it. 603 00:36:59,840 --> 00:37:01,640 Also down 604 00:37:01,640 --> 00:37:06,920 below, I calculated with the dataset from, from Lausanne, from Switzerland, 605 00:37:06,920 --> 00:37:10,920 with young people watching movie all the time. 606 00:37:11,000 --> 00:37:15,320 So I started slowly to be convinced is really smaller. 607 00:37:15,400 --> 00:37:17,920 And then I got the interpretation. 608 00:37:17,920 --> 00:37:20,960 I mean, well, the first interpretation is, 609 00:37:20,960 --> 00:37:25,480 finally demonstrate why movies are so relaxing 610 00:37:25,560 --> 00:37:29,040 is because they are disconnecting the brain. 611 00:37:29,120 --> 00:37:31,520 But actually the truest nature. 612 00:37:31,520 --> 00:37:35,400 Of course, movies are relaxing, I think is the word 613 00:37:35,520 --> 00:37:38,720 resting state is absolutely misleading. 614 00:37:38,800 --> 00:37:43,320 Of course, during resting state we are doing much more things that we 615 00:37:43,400 --> 00:37:46,560 that that would that 616 00:37:46,680 --> 00:37:51,520 we are executing in movie watching. 617 00:37:51,600 --> 00:37:56,400 And it's confusing the word resting I mean that implicitly. 618 00:37:56,480 --> 00:37:56,960 Okay. 619 00:37:56,960 --> 00:38:00,200 So we have very very clear quantitative results. 620 00:38:00,200 --> 00:38:03,680 So we create a model in this case is one model 621 00:38:03,680 --> 00:38:07,280 for each single participant in each single condition. 622 00:38:07,280 --> 00:38:14,040 So one model for participant one watch in a movie or resting state. 623 00:38:14,120 --> 00:38:16,400 And then we take the GIC. 624 00:38:16,400 --> 00:38:19,600 So the generative affective connectivity and we try to prove, 625 00:38:19,600 --> 00:38:22,600 okay, this generative effect, the connectivity, 626 00:38:22,800 --> 00:38:26,400 the generators of the Iraqi, is the mechanistic explanation 627 00:38:26,400 --> 00:38:29,560 of the generation of the, you know, should be useful. 628 00:38:29,720 --> 00:38:34,440 How I showed that while the easiest way is let's try to classify 629 00:38:34,640 --> 00:38:39,960 so we use standard classifiers in this case support vector machine, 630 00:38:40,040 --> 00:38:44,600 for example, we try to classify the conditions 631 00:38:44,600 --> 00:38:49,720 if one particular person is watching a movie or is in rest. 632 00:38:49,800 --> 00:38:51,200 And as you see 633 00:38:51,200 --> 00:38:55,160 on the right, I mean, or just get the feeling by the box plot, 634 00:38:55,160 --> 00:39:00,600 which are very high, the level of accuracy of these classifiers is pretty high. 635 00:39:00,680 --> 00:39:04,320 So over 90% if you do with the functional connectivity. 636 00:39:04,320 --> 00:39:09,440 So forget thermodynamics, forget modeling, forget everything, just 637 00:39:09,520 --> 00:39:13,240 the absolutely stupid functional connectivity of your data. 638 00:39:13,440 --> 00:39:16,800 So just a correlation you doing also good job 639 00:39:16,880 --> 00:39:20,840 is to is a blue box blood a but not always. 640 00:39:21,000 --> 00:39:24,720 The conditions where you are practically just labor in particular 641 00:39:24,720 --> 00:39:29,160 is what you see right bottom. 642 00:39:29,240 --> 00:39:31,800 We classify there are two type of movies 643 00:39:31,800 --> 00:39:34,520 Hollywood movies and creative common 644 00:39:34,520 --> 00:39:38,600 movies, Creative Commons movies, 645 00:39:38,680 --> 00:39:40,640 YouTubers, things like that. 646 00:39:40,640 --> 00:39:43,040 And Hollywood are the Hollywood movies are 647 00:39:43,040 --> 00:39:48,600 Mission Impossible, James Bond, as all the 648 00:39:48,680 --> 00:39:53,400 A F sees, is very bad in classifying that. 649 00:39:53,480 --> 00:39:55,040 But the Jeep. 650 00:39:55,040 --> 00:39:58,240 So the generators of the you don't go when decision is over 651 00:39:58,240 --> 00:40:02,320 90%, you can distinguish between the different type of movies. 652 00:40:02,400 --> 00:40:06,160 I would love to repeat that between Hollywood and European movies. 653 00:40:06,160 --> 00:40:09,160 But then I don't have to do it. 654 00:40:09,320 --> 00:40:11,440 I am sure that that defense 655 00:40:11,440 --> 00:40:15,360 same technology model base 656 00:40:15,440 --> 00:40:17,640 of the inside, out of the you are going imitation 657 00:40:17,640 --> 00:40:19,760 for the same dataset that I was showing before. 658 00:40:19,760 --> 00:40:24,360 We then it the biomedical dataset, the coma cases 659 00:40:24,440 --> 00:40:29,280 and doing the same of course in the box lot the is not shown here. 660 00:40:29,280 --> 00:40:32,320 The difference is like 661 00:40:32,320 --> 00:40:35,400 internet in the level of unionization. 662 00:40:35,400 --> 00:40:39,240 But when you do the model and you try to classify the model 663 00:40:39,320 --> 00:40:42,840 in controls minimal consciousness or uncommunicative 664 00:40:43,000 --> 00:40:46,720 wakefulness in drones or deep, deep coma, 665 00:40:46,800 --> 00:40:51,320 then you have a very good level of of separation. 666 00:40:51,400 --> 00:40:55,360 And if you want of of diagnosing 667 00:40:55,400 --> 00:41:01,600 really the type of stage of coma that that you have in that condition, 668 00:41:01,600 --> 00:41:06,160 which is of course a hot topic in in in coma research. 669 00:41:06,240 --> 00:41:09,480 I just want to finish with some 670 00:41:09,480 --> 00:41:13,160 speculations about the future, 671 00:41:13,240 --> 00:41:16,240 how we can far that apply 672 00:41:16,400 --> 00:41:20,360 really this these ideas of thermodynamics 673 00:41:20,360 --> 00:41:26,440 and now in a much more sophisticated way to get extra information 674 00:41:26,520 --> 00:41:29,800 in just based on the same conceptual idea. 675 00:41:29,800 --> 00:41:31,720 I mean in equilibrium 676 00:41:31,720 --> 00:41:37,560 we have no knowledge of time ever seen, not reversibility and so on. 677 00:41:37,640 --> 00:41:40,080 And in non-equilibrium we have something different. 678 00:41:40,080 --> 00:41:43,080 We have to give quantization to 679 00:41:43,160 --> 00:41:47,280 A and that is a theorem which if you are not from physics, 680 00:41:47,280 --> 00:41:52,360 even if you are from physics, I forgot that theorem. 681 00:41:52,360 --> 00:41:54,800 I mean, to be honest with you, when I started to look that again, 682 00:41:54,800 --> 00:41:56,840 it's a fluctuation dissipation theorem. 683 00:41:56,840 --> 00:42:00,640 Scardoelli mean, when you say flotation of what this equation of what? 684 00:42:00,720 --> 00:42:03,400 Well, that was the first 685 00:42:03,400 --> 00:42:08,000 important contribution of Albert Einstein 686 00:42:08,080 --> 00:42:10,880 when he was in Ban at the patent office. 687 00:42:10,880 --> 00:42:15,200 Doing bureaucratic work seems to be that he was extremely wor 688 00:42:15,280 --> 00:42:18,040 and he started to look at the 689 00:42:18,040 --> 00:42:21,480 what we call nowadays the Brownian movement. 690 00:42:21,560 --> 00:42:25,680 So you put in a glass of water, little the particles, 691 00:42:25,760 --> 00:42:30,840 and then he was really just looking at how they fluctuating fluctuation. 692 00:42:30,920 --> 00:42:32,800 And he explained fact that 693 00:42:32,800 --> 00:42:37,200 what happens as he was really 694 00:42:37,280 --> 00:42:40,760 very bored and then say what do what should they do now? 695 00:42:40,760 --> 00:42:42,840 Okay, let's put some charge on the particle. 696 00:42:42,840 --> 00:42:44,360 And I put an electric field. 697 00:42:44,360 --> 00:42:46,360 So I produced this system. 698 00:42:46,360 --> 00:42:49,000 And what he found is fabulous 699 00:42:49,000 --> 00:42:52,000 and is the and this is the dissipation part. 700 00:42:52,160 --> 00:42:57,320 Namely, that's a if the system is in equilibrium, 701 00:42:57,400 --> 00:43:01,400 the fluctuation, but addict hundred percent, 702 00:43:01,480 --> 00:43:06,440 the dissipation that is magic is not magic, of course. 703 00:43:06,520 --> 00:43:08,280 But but it sounds like magic. 704 00:43:08,280 --> 00:43:14,080 I mean, I just look how they fluctuate when I not produce the system. 705 00:43:14,160 --> 00:43:16,600 And that is enough. 706 00:43:16,600 --> 00:43:17,920 I can predict Perfect. 707 00:43:17,920 --> 00:43:18,880 I don't need to measure. 708 00:43:18,880 --> 00:43:22,480 I can predict what would happen when you party of the system. 709 00:43:22,560 --> 00:43:24,960 Fantastic 710 00:43:25,040 --> 00:43:26,120 parenthesis. 711 00:43:26,120 --> 00:43:28,520 Let's go to neuroscience, 712 00:43:28,520 --> 00:43:31,040 one of my essays in energy Partners. 713 00:43:31,040 --> 00:43:32,320 But it's not because of that. 714 00:43:32,320 --> 00:43:35,520 I discuss all that they got into that. 715 00:43:35,600 --> 00:43:39,160 And with this Marcelo, my cimini 716 00:43:39,240 --> 00:43:43,880 he had is a medical doctor, had a fabulous intuition and a fabulous idea. 717 00:43:43,880 --> 00:43:49,320 He has no idea about Albert Einstein, has no idea about the this patient. 718 00:43:49,400 --> 00:43:53,080 And many years ago they have this great idea, 719 00:43:53,080 --> 00:43:56,840 we will produce the brain with DMEs, for example. 720 00:43:56,880 --> 00:43:59,960 That's the easiest way I will characterize the brain 721 00:43:59,960 --> 00:44:02,960 in a in a very cheap way, e.g. 722 00:44:03,080 --> 00:44:05,320 so really cheap. 723 00:44:05,320 --> 00:44:08,000 And he had this intuition. 724 00:44:08,000 --> 00:44:11,720 That's when a part of the brain and they have different 725 00:44:11,720 --> 00:44:14,080 underlying dynamics because it's associated 726 00:44:14,080 --> 00:44:19,440 with different brain states, sleep, anesthesia, coma, whatever. 727 00:44:19,520 --> 00:44:21,240 Then the effect of 728 00:44:21,240 --> 00:44:24,240 the perturbation is different 729 00:44:24,280 --> 00:44:27,800 and then say how a measure that I have no idea. 730 00:44:27,800 --> 00:44:29,920 I never studied the physics or I have. 731 00:44:29,920 --> 00:44:34,360 So I used a more simple thing, which is fantastic. 732 00:44:34,440 --> 00:44:36,120 And this is what he called the PCI. 733 00:44:36,120 --> 00:44:40,080 The perturbation Complexity index are used to compress the ability. 734 00:44:40,080 --> 00:44:43,760 So the sleep that you have on your laptop, you don't need to understand 735 00:44:43,760 --> 00:44:47,360 even celibacy for complexity. 736 00:44:47,440 --> 00:44:49,280 So I use impulsive complexity. 737 00:44:49,280 --> 00:44:52,280 So I think how much I can compress 738 00:44:52,480 --> 00:44:56,520 the signal that I evoke after the perturbation 739 00:44:56,600 --> 00:44:59,840 and they managed to publish 740 00:44:59,920 --> 00:45:06,040 very relevant papers, nature signs and in many other journals 741 00:45:06,120 --> 00:45:09,320 showing basically what is the nice, what is not gotten. 742 00:45:09,320 --> 00:45:14,720 I too, from one of the early papers that's under different conditions like 743 00:45:14,800 --> 00:45:18,000 resting wakefulness or sleep 744 00:45:18,080 --> 00:45:21,640 and non-REM sleep or REM sleep or coma. 745 00:45:21,640 --> 00:45:25,520 What you see at the bottom right again, deep coma, 746 00:45:25,520 --> 00:45:29,600 meaning out of consciousness or anesthesia is different here. 747 00:45:29,640 --> 00:45:33,400 You see, even without the PCI, I am not the blood in the box, 748 00:45:33,400 --> 00:45:36,880 blood of the PCI hip. Lots to that, of course. 749 00:45:36,960 --> 00:45:37,280 But you 750 00:45:37,280 --> 00:45:41,000 see here just by the cartoon that the signal is simple 751 00:45:41,000 --> 00:45:46,360 and therefore is more compressible, the more unconscious you are. 752 00:45:46,440 --> 00:45:48,320 A fantastic idea. 753 00:45:48,320 --> 00:45:51,480 Really fantastic idea. Fantastic intuition. 754 00:45:51,560 --> 00:45:54,840 And they, they and they, of course, develop it 755 00:45:54,840 --> 00:45:57,800 that in the context of not only of of biomedical application, 756 00:45:57,800 --> 00:46:02,280 but in the context of consciousness research. 757 00:46:02,360 --> 00:46:04,520 Now I try to link and of the parenthesis 758 00:46:04,520 --> 00:46:08,240 I try to link the fluctuation dissipates in theorem 759 00:46:08,320 --> 00:46:12,880 with this and with the more dynamics and this is how I will finish. 760 00:46:12,960 --> 00:46:15,440 The idea is very simple and 761 00:46:15,440 --> 00:46:19,280 I hope that you already came to the idea 762 00:46:19,360 --> 00:46:22,360 that works because the system is non-equilibrium. 763 00:46:22,600 --> 00:46:25,080 If it would be equilibrium, 764 00:46:25,080 --> 00:46:28,080 then a massive mini 765 00:46:28,200 --> 00:46:32,640 would have discovered that I don't learn anything 766 00:46:32,720 --> 00:46:35,120 because I could distinguish them of course, 767 00:46:35,120 --> 00:46:36,800 but I do not need the perturbation. 768 00:46:36,800 --> 00:46:40,640 I just look at the fluctuation level and that should be enough. 769 00:46:40,720 --> 00:46:42,440 But it's not enough. Why? 770 00:46:42,440 --> 00:46:45,440 Because all those brains are in Non-Equilibrium 771 00:46:45,440 --> 00:46:50,160 and therefore the fluctuation dissipation theorem does not hold. 772 00:46:50,240 --> 00:46:55,000 And that means that the degree of violation of the fluctuation 773 00:46:55,000 --> 00:46:59,680 dissipation theorem or the degree of hierarchy, realization 774 00:46:59,680 --> 00:47:06,480 of the degree of arrow time, or the degree of non-equilibrium 775 00:47:06,560 --> 00:47:09,720 is a measure of the brain of states. 776 00:47:09,800 --> 00:47:14,880 And instead now of characterizing this brain state with data of time, 777 00:47:14,880 --> 00:47:18,680 as we have done during the whole talk, in all the different styles, model, 778 00:47:18,680 --> 00:47:22,800 free model based, I would use perturbation, 779 00:47:22,880 --> 00:47:27,000 and that is a beautiful way of doing that. 780 00:47:27,080 --> 00:47:30,720 So if the brain would be in equilibrium, what your is 781 00:47:30,800 --> 00:47:34,680 and I say to you that the spontaneous state, the resting state, 782 00:47:34,760 --> 00:47:38,360 and this mathematically that I see it in a very simple way. 783 00:47:38,360 --> 00:47:41,480 I mean, I don't go into the details, but it's just correlations. 784 00:47:41,480 --> 00:47:44,960 It's like the vanishing connectivity and the dissipation 785 00:47:44,960 --> 00:47:49,320 which is basically what we call susceptibility is the effect of the. 786 00:47:49,400 --> 00:47:53,200 So it's a more sophisticated version of the level of 787 00:47:53,280 --> 00:47:54,480 they should be equal. 788 00:47:54,480 --> 00:47:57,240 And that is the fluctuation dissipation theorem. 789 00:47:57,240 --> 00:48:00,760 But that is not the case because the brain is in non-equilibrium. 790 00:48:00,840 --> 00:48:04,040 And therefore what we want to see is 791 00:48:04,120 --> 00:48:07,440 how much we are contradicting this equality. 792 00:48:07,560 --> 00:48:10,960 So we define the measure, which is basically the difference 793 00:48:11,120 --> 00:48:16,080 of the left and right hand side that normalize it. 794 00:48:16,160 --> 00:48:21,480 And this is the degree of deviation from the fluctuation dissipation theorem. 795 00:48:21,560 --> 00:48:24,320 And this is a measure, another measure, 796 00:48:24,320 --> 00:48:29,080 a clever measure, because it's used in something that is not the in in the 797 00:48:29,080 --> 00:48:30,200 in the pure fluctuations, 798 00:48:30,200 --> 00:48:34,600 you are adding the defect of the dissipation of the perturbation. 799 00:48:34,680 --> 00:48:37,440 And of course, one can do this 800 00:48:37,440 --> 00:48:41,040 with the model exactly the same as before. 801 00:48:41,120 --> 00:48:45,640 And you produced the model in Silico in all these different way 802 00:48:45,640 --> 00:48:49,440 and just measure different anticipation deviation. 803 00:48:49,520 --> 00:48:52,320 And what you see here, just look at the box below. 804 00:48:52,320 --> 00:48:56,480 I mean, the renderings is that this is just about the we did, 805 00:48:56,560 --> 00:48:59,960 but just in the box plot, you see two cases 806 00:48:59,960 --> 00:49:03,720 sleep against Awake human neuroimaging 807 00:49:03,800 --> 00:49:07,400 small set 18 participant at the date of hand was love 808 00:49:07,600 --> 00:49:11,200 very well known and the data that we use it all the time? 809 00:49:11,200 --> 00:49:16,360 HCB More or less thousand participants in all this condition repressing cognition. 810 00:49:16,360 --> 00:49:20,360 So and the degree of violation of the voltage anticipation theorem 811 00:49:20,400 --> 00:49:24,720 is a good quantitative measure of the degree of non nicotine, 812 00:49:24,800 --> 00:49:27,560 so that you are justifying 813 00:49:27,560 --> 00:49:32,640 why the PCI works and how you 814 00:49:32,720 --> 00:49:34,640 can define the. 815 00:49:34,640 --> 00:49:39,640 It sounds a little bit arrogant, but it's not that the right version of PCI 816 00:49:39,640 --> 00:49:43,440 because it's not intuitive. 817 00:49:43,440 --> 00:49:46,440 It's the one that came in from first principles, 818 00:49:46,440 --> 00:49:49,840 namely the fluctuation, dissipation, delay and in that direction 819 00:49:49,920 --> 00:49:51,200 that we are trying to go. 820 00:49:51,200 --> 00:49:56,000 Now, they say I think they go very rapidly on the conclusion 821 00:49:56,000 --> 00:50:00,880 their main objective was hierarchical organization, 822 00:50:00,960 --> 00:50:04,760 corporate level, using whatever you use e.g. 823 00:50:04,760 --> 00:50:08,840 me, G, I and I, even local field potentials 824 00:50:08,840 --> 00:50:12,680 animals, humans in all possible conditions. 825 00:50:12,760 --> 00:50:14,800 And the trick was extremely simple. 826 00:50:14,800 --> 00:50:17,880 It just looked out of time with different methodologies, with machine 827 00:50:17,880 --> 00:50:22,080 learning, with correlations. 828 00:50:22,160 --> 00:50:25,720 so basically that is the main they commit out of time. 829 00:50:25,720 --> 00:50:29,280 Non-Equilibrium And you don't go any stage on the word turbulence. 830 00:50:29,280 --> 00:50:31,400 Forget it comes around. I did talk of that. 831 00:50:31,400 --> 00:50:35,800 I'm not mentioning today, but this is strongly related with that. 832 00:50:35,880 --> 00:50:39,640 And we have done a model of that and we proved that the generators know 833 00:50:39,640 --> 00:50:43,440 the parameters of that model, have efficient interpretation of these 834 00:50:43,440 --> 00:50:48,120 what we call generative effect. Connectivity is informative. 835 00:50:48,200 --> 00:50:49,440 Is that 836 00:50:49,440 --> 00:50:53,240 then something about the origin of this hierarchical organization? 837 00:50:53,240 --> 00:50:58,240 And nowadays we are trying to develop a sophisticated, 838 00:50:58,320 --> 00:51:02,880 sophisticated technique in the framework of the fluctuation dissipation data. 839 00:51:02,960 --> 00:51:07,840 So I was doing all this with my good friend from Oxford, Morten Kringle. 840 00:51:07,840 --> 00:51:13,520 Why we develop it, all these ideas, basically, do we think of it? 841 00:51:13,600 --> 00:51:15,960 We were also born, not invented. 842 00:51:15,960 --> 00:51:17,760 This at home. 843 00:51:17,760 --> 00:51:23,280 Was that really a pleasure to work with him online on all these things? 844 00:51:23,360 --> 00:51:32,440 So thank you very much for your attention. 845 00:51:32,520 --> 00:51:36,680 But wow, that's a lot of food 846 00:51:36,680 --> 00:51:42,320 for thought, Gustavo, and that I have a lot of questions that I want. 847 00:51:42,520 --> 00:51:44,040 One point. 848 00:51:44,040 --> 00:51:48,320 So you derive two indices, one 849 00:51:48,320 --> 00:51:53,000 through the reversibility component and then another one when you compare 850 00:51:53,080 --> 00:51:55,560 functional conductivity with this more 851 00:51:55,560 --> 00:51:59,480 elegant dissipation equation. 852 00:51:59,560 --> 00:52:02,200 And what I found interesting is the ranking 853 00:52:02,200 --> 00:52:04,360 of the cognitive states from the Human Connection 854 00:52:04,360 --> 00:52:07,600 Project is the same with the both techniques, isn't it? 855 00:52:07,680 --> 00:52:09,320 It's roughly the same. Yeah. 856 00:52:09,320 --> 00:52:13,200 So you said the extremes are the same social and emotional, such as 857 00:52:13,200 --> 00:52:16,360 is the maximum, the emotion is the minimum them in the squid. 858 00:52:16,360 --> 00:52:17,440 Interesting to me. 859 00:52:17,440 --> 00:52:20,360 So there's something tapping into 860 00:52:20,360 --> 00:52:22,080 something that you use. 861 00:52:22,080 --> 00:52:26,600 It's yes, I mean the the actual one 862 00:52:26,600 --> 00:52:29,600 can one can look really 863 00:52:29,600 --> 00:52:31,760 when can look the renderings 864 00:52:31,760 --> 00:52:37,160 of the local version of the arrow of time and then you see that that's really there 865 00:52:37,160 --> 00:52:42,960 much more asymmetry in a complex task like the social task of the HTP 866 00:52:43,040 --> 00:52:47,080 than in the emotional task, which is basically just visual perception. 867 00:52:47,080 --> 00:52:50,320 I mean, they call emotion because they distinguish as angry 868 00:52:50,360 --> 00:52:53,360 nothing but the that there is an emotional component. 869 00:52:53,360 --> 00:52:56,320 You see really the typical suspects. 870 00:52:56,320 --> 00:52:59,760 I mean that there are asymmetries in the interactions 871 00:52:59,840 --> 00:53:02,600 but that there is a much more massive 872 00:53:02,600 --> 00:53:06,640 asymmetries, relational or organized orchestration 873 00:53:06,640 --> 00:53:11,480 of computation in complex task than in simple task, as I suspect. 874 00:53:11,560 --> 00:53:13,360 And one thing that I didn't 875 00:53:13,360 --> 00:53:17,640 quite understand is how you do the perturbation 876 00:53:17,720 --> 00:53:20,360 on the last part. Yeah, yeah. 877 00:53:20,360 --> 00:53:22,480 So we take the model, 878 00:53:22,480 --> 00:53:27,480 individual participant, individual condition, 879 00:53:27,560 --> 00:53:31,800 we take the model in Silico and then we produce the model. 880 00:53:31,880 --> 00:53:35,720 So each single note is perturb it in some way. 881 00:53:35,720 --> 00:53:36,960 I inject noise. 882 00:53:36,960 --> 00:53:40,680 All right, Inject noise beginning right over the air, the beauty 883 00:53:40,720 --> 00:53:44,400 that you can do some analytical tricks because the model is linear. 884 00:53:44,400 --> 00:53:45,200 I see it. 885 00:53:45,200 --> 00:53:49,560 And and, you know, at the end of it, you have an analytical version 886 00:53:49,640 --> 00:53:51,280 of the dissipation. 887 00:53:51,280 --> 00:53:53,600 So therefore it is very rapid to calculate. 888 00:53:53,600 --> 00:53:57,360 Okay, You know, I was thinking it had to be knock out a node or inject something. 889 00:53:57,360 --> 00:53:58,240 Okay. 890 00:53:58,240 --> 00:54:04,120 And just one of the thing I want one point and with a lot of MRI function 891 00:54:04,120 --> 00:54:08,400 imaging measures is a single patient data 892 00:54:08,400 --> 00:54:13,000 is always very noisy to to classify a single patient as schizophrenia. 893 00:54:13,000 --> 00:54:14,440 You have one event. 894 00:54:14,440 --> 00:54:18,680 But I just wondering whether your global measure 895 00:54:18,760 --> 00:54:21,680 should the global measure of non reversibility 896 00:54:21,680 --> 00:54:25,120 for example how well that would work on an individual. 897 00:54:25,120 --> 00:54:27,120 Yeah. 898 00:54:27,200 --> 00:54:30,320 Indirectly I saw some results. 899 00:54:30,320 --> 00:54:35,360 So the schizophrenia result bipolar ADHD or the coma results, 900 00:54:35,600 --> 00:54:38,680 there were single single patients. Yeah. 901 00:54:38,800 --> 00:54:40,120 The dots were single patients. 902 00:54:40,120 --> 00:54:40,480 Yeah. 903 00:54:40,480 --> 00:54:45,800 All, all the analysis, even model for your model based single patients 904 00:54:45,880 --> 00:54:49,760 and the degree of and using a global measure you can use. 905 00:54:49,760 --> 00:54:56,280 Well the that is is pretty informative so the effective connectivity is very good 906 00:54:56,360 --> 00:55:00,200 but just a global measure of non-equilibrium in some case is very good 907 00:55:00,200 --> 00:55:01,120 to distinguish it. 908 00:55:01,120 --> 00:55:05,440 It's not so good for classifying of course is to course to block it. 909 00:55:05,520 --> 00:55:08,720 But if you go into the details and for example, 910 00:55:08,800 --> 00:55:12,120 one example is the cheek, the affective connectivity 911 00:55:12,200 --> 00:55:17,120 done, done the label of classification article and really high and really high. 912 00:55:17,120 --> 00:55:21,760 So that so that even with the with poor data, because the quality 913 00:55:21,760 --> 00:55:26,160 of the data are really not to encompass, you know, the neuroimaging. 914 00:55:26,160 --> 00:55:30,720 I mean these are next to the factor and if that's the brain is 915 00:55:30,800 --> 00:55:32,720 pretty distorted, 916 00:55:32,720 --> 00:55:36,160 Yeah, but is, is another source of noise. 917 00:55:36,200 --> 00:55:40,920 I mean that each individual is a nightmare. 918 00:55:41,000 --> 00:55:43,480 But in spite of that that you saw 919 00:55:43,480 --> 00:55:47,520 I think was over 80% the classification. 920 00:55:47,600 --> 00:55:47,840 Yeah. 921 00:55:47,840 --> 00:55:49,960 And that's what I was saying that was doing very well. 922 00:55:49,960 --> 00:55:54,000 I mean, all of these diseases as well, they share the same genetics 923 00:55:54,080 --> 00:55:57,480 and say schizophrenia and bipolar disorder and ADHD. 924 00:55:57,480 --> 00:56:02,720 So It would be surprising that you differentiate well between them 925 00:56:02,760 --> 00:56:04,720 and have the same functional networks as well. 926 00:56:04,720 --> 00:56:09,560 So that's why it struck me that it does it does well as a technique. 927 00:56:09,640 --> 00:56:13,960 Yeah, we didn't try to classify it, to be honest with you. 928 00:56:13,960 --> 00:56:17,400 The the UCLA, 929 00:56:17,480 --> 00:56:19,800 but we classify 930 00:56:19,800 --> 00:56:24,120 other psychiatric diseases or even smaller, for example, 931 00:56:24,320 --> 00:56:28,920 the depression that I said from robot Robin Carhart-Harris, 932 00:56:29,000 --> 00:56:32,000 the one that we actually use psychedelics. 933 00:56:32,000 --> 00:56:37,200 And so it's a beautiful thing because he uses citalopram that 934 00:56:37,280 --> 00:56:40,480 the traditional they said I 935 00:56:40,560 --> 00:56:44,400 drug for depression or the psychedelics 936 00:56:44,480 --> 00:56:48,600 and then you have both them and then you can see the effects of loss 937 00:56:48,680 --> 00:56:51,960 and it says model that I said because they have more 938 00:56:51,960 --> 00:56:55,840 or less 20 patients bad 939 00:56:55,920 --> 00:56:59,000 condition and that works very good. 940 00:56:59,120 --> 00:57:04,000 I mean, the classifier of we do classify it's not Jake is a little bit different. 941 00:57:04,000 --> 00:57:09,080 Image is another characterization of the hierarchy but not global. 942 00:57:09,080 --> 00:57:13,680 It's much more detailed is what we call trophic characterization 943 00:57:13,760 --> 00:57:18,200 and that classify it relatively well even can predict 944 00:57:18,280 --> 00:57:21,840 the response to the drug, which is it for me is the Holy Grail. 945 00:57:21,920 --> 00:57:23,760 Yeah. And so that's going to be looked for. 946 00:57:23,760 --> 00:57:27,760 So a lot of these things we don't have that much clinical applications 947 00:57:27,760 --> 00:57:31,520 for animal research despite the huge amount of MRI research 948 00:57:31,520 --> 00:57:35,120 that we had just squeakers for this one issue that is not so reliable 949 00:57:35,360 --> 00:57:40,760 on an individual patient basis, except okay, 950 00:57:40,840 --> 00:57:42,640 I have more questions, but I'm being greedy. 951 00:57:42,640 --> 00:57:46,760 So someone else went or something. 952 00:57:46,840 --> 00:57:49,200 Yeah. 953 00:57:49,200 --> 00:57:50,440 So thank you 954 00:57:50,440 --> 00:57:53,000 for this is an interesting topic 955 00:57:53,000 --> 00:57:55,840 and I would like to know your opinion 956 00:57:55,840 --> 00:57:59,960 about the 957 00:58:00,040 --> 00:58:04,640 the causality for all these type of molars. 958 00:58:04,640 --> 00:58:08,360 At the beginning you mentioned 959 00:58:08,440 --> 00:58:09,760 causality, 960 00:58:09,760 --> 00:58:12,320 but it seems to be 961 00:58:12,320 --> 00:58:15,320 not enough for several situations. 962 00:58:15,400 --> 00:58:18,360 So I don't know if you are familiar 963 00:58:18,360 --> 00:58:21,640 with the type of alternatives for 964 00:58:21,720 --> 00:58:24,760 so on causality for measured in the study. 965 00:58:25,000 --> 00:58:27,520 Yeah. Yeah. 966 00:58:27,520 --> 00:58:31,440 I don't know. And 967 00:58:31,520 --> 00:58:36,360 in fact I'm in the second version of the model for the inside out. 968 00:58:36,440 --> 00:58:41,560 It's a cheap version of the Granger Causality 969 00:58:41,640 --> 00:58:44,880 A and, and that was enough 970 00:58:44,960 --> 00:58:47,600 but things because we are not so much interested 971 00:58:47,600 --> 00:58:51,600 in the absolute value I mean perhaps this body but in characterizing 972 00:58:51,680 --> 00:58:57,360 the degree of interactions, if you are really interested on the values 973 00:58:57,440 --> 00:59:01,920 but is enough to assess the degree of asymmetry. 974 00:59:02,000 --> 00:59:05,200 So my my, 975 00:59:05,280 --> 00:59:10,280 my, my belief or my intuition nowadays is that thanks to this 976 00:59:10,280 --> 00:59:15,880 thermodynamic group, we are allowed to measure causality in a bad way. 977 00:59:15,960 --> 00:59:18,080 And that's would be enough, 978 00:59:18,080 --> 00:59:22,760 of course, if I if I would be able to find something 979 00:59:22,760 --> 00:59:28,960 better than the Granger causality or transfer entropy, 980 00:59:29,040 --> 00:59:30,280 I would use that. 981 00:59:30,280 --> 00:59:33,200 And if I have enough data, I would use that 982 00:59:33,200 --> 00:59:35,720 and I would then do the thermodynamic trick. 983 00:59:35,720 --> 00:59:40,800 I just would construct the graph and analyze in the traditional way. 984 00:59:40,880 --> 00:59:44,720 But I was not able to find that way for 985 00:59:44,880 --> 00:59:49,320 and robust enough for applying this to to individual patients. 986 00:59:49,400 --> 00:59:53,520 We have a really very bad quality of data, and this is 987 00:59:53,600 --> 01:00:01,920 that in 988 01:00:02,000 --> 01:00:03,600 that the example 989 01:00:03,600 --> 01:00:07,960 you gave of this reversibility with the Russian name 990 01:00:08,040 --> 01:00:15,120 at the beginning, what was the name of the reversibility of movies, 991 01:00:15,200 --> 01:00:20,320 the reversibility of movies in time? 992 01:00:20,400 --> 01:00:22,360 Are the 993 01:00:22,360 --> 01:00:24,880 the movie, you mean? Yeah. 994 01:00:24,880 --> 01:00:26,240 Christopher Nolan No, no, no. 995 01:00:26,240 --> 01:00:28,120 That, that 996 01:00:28,120 --> 01:00:29,720 someone that had shown that it was reversible, 997 01:00:29,720 --> 01:00:33,120 but over a small timescale just since, just introducing, you know. 998 01:00:33,120 --> 01:00:37,480 So what sort of timescales are we thinking of and what, 999 01:00:37,480 --> 01:00:42,320 when when they were, when they come up with this idea 1000 01:00:42,400 --> 01:00:45,760 that was and it's been actually 1001 01:00:45,840 --> 01:00:51,000 it's artificial timescales, I mean but the if, if you would apply 1002 01:00:51,000 --> 01:00:56,440 these are exactly the same type of spin model that we use in zero image. 1003 01:00:56,440 --> 01:00:58,880 And by the way from time to time. 1004 01:00:58,880 --> 01:01:01,880 So you can apply these 1005 01:01:02,000 --> 01:01:05,000 if you want, even at the millisecond scale if you want. 1006 01:01:05,160 --> 01:01:05,520 Yeah. 1007 01:01:05,520 --> 01:01:06,880 Yeah, I thought so. 1008 01:01:06,880 --> 01:01:13,160 But I think what you are trying to, 1009 01:01:13,240 --> 01:01:14,040 to formulate 1010 01:01:14,040 --> 01:01:17,960 is a good question that I don't know the answer, 1011 01:01:17,960 --> 01:01:22,560 but they will work against me and is actually is a 1012 01:01:22,600 --> 01:01:26,360 thin is a challenging question if the level of non-equilibrium 1013 01:01:26,360 --> 01:01:31,320 or if the level of communication changed with the timescale 1014 01:01:31,400 --> 01:01:35,600 that I think I believe that of course I don't know. 1015 01:01:35,720 --> 01:01:38,520 We never highlighted one can do that. 1016 01:01:38,520 --> 01:01:39,840 We have not done that. 1017 01:01:39,840 --> 01:01:42,440 For example, in the manga or with imaging. 1018 01:01:42,440 --> 01:01:46,280 And then you really on purpose, you, you just concentrate 1019 01:01:46,280 --> 01:01:50,320 on different bands or different filters and then you have the answer. 1020 01:01:50,400 --> 01:01:52,720 Yeah, that is we have never done that. 1021 01:01:52,720 --> 01:01:55,160 It's a good question about that. Yeah, but there's also a question. 1022 01:01:55,160 --> 01:02:01,360 So we in electrophysiology we use Granger calls on a team and, and 1023 01:02:01,440 --> 01:02:02,520 sometimes review. 1024 01:02:02,520 --> 01:02:06,120 I sometimes happen to say have you, have you, have you flip the data 1025 01:02:06,200 --> 01:02:10,240 around to make sure that there is no Granger causality. 1026 01:02:10,320 --> 01:02:13,320 that should be a good answer goes I did but the thing so 1027 01:02:13,520 --> 01:02:17,960 yeah the because basically showing that information flows in one direction 1028 01:02:17,960 --> 01:02:20,360 so the history of area predicts the history of area 1029 01:02:20,360 --> 01:02:24,240 B better than the history B itself on its own. 1030 01:02:24,320 --> 01:02:25,640 Yeah I think I got that right. 1031 01:02:25,640 --> 01:02:29,640 Anyway so just sort of regressive things and, 1032 01:02:29,720 --> 01:02:33,320 and there should be and that's what I was going 1033 01:02:33,400 --> 01:02:37,560 I was wondering okay so maybe, you know, you can be lucky 1034 01:02:37,640 --> 01:02:40,520 and if you're doing with higher frequencies, you get one result. 1035 01:02:40,520 --> 01:02:43,520 Then if you're dealing with low frequencies, when you flip the. 1036 01:02:43,600 --> 01:02:44,960 Absolutely. 1037 01:02:44,960 --> 01:02:47,520 Actually we have that of one of our reviewers also 1038 01:02:47,520 --> 01:02:53,480 was asking, well, you are telling that thing that in a to cartoonish or it 1039 01:02:53,480 --> 01:02:58,040 depends on causality similar to the cat you showed that and they have shown that 1040 01:02:58,040 --> 01:03:03,480 I mean I took the ACP I calculated I had operated like an engine goes I did 1041 01:03:03,560 --> 01:03:07,840 I calculated the degree of asymmetry of the Granger I was head of the classical 1042 01:03:07,840 --> 01:03:13,760 Granger causality A and that correlates perfectly with the inside out. 1043 01:03:13,840 --> 01:03:16,280 All right. Okay. Well, that's good to know. 1044 01:03:16,280 --> 01:03:21,040 Yeah, it's a I was annoyed because it was extra work, but 1045 01:03:21,120 --> 01:03:21,440 then we 1046 01:03:21,440 --> 01:03:25,240 can refer to it and to review of comments, please see Decker's 1047 01:03:25,320 --> 01:03:28,000 response on this reversibility of Granger causality. 1048 01:03:28,000 --> 01:03:32,520 Yeah, there's another point here and that would then work in your machine. 1049 01:03:32,520 --> 01:03:37,120 But and there's a lot of debate as to how much the structure, 1050 01:03:37,120 --> 01:03:41,560 the function of connectivity in the brain matches the structural connectivity. 1051 01:03:41,640 --> 01:03:46,800 So there is some gross component that and you know, the areas that have bigger 1052 01:03:46,800 --> 01:03:49,920 connections will be fluctuation, the blood 1053 01:03:49,920 --> 01:03:53,120 oxygen level up and down a bit more. 1054 01:03:53,200 --> 01:03:55,280 But you go one step further with your modeling. 1055 01:03:55,280 --> 01:03:59,680 So you're absolutely you use 1056 01:03:59,760 --> 01:04:02,160 this the generative effect of connectivity 1057 01:04:02,160 --> 01:04:05,160 defined by the structural connectivity. 1058 01:04:05,240 --> 01:04:06,000 Yeah. 1059 01:04:06,000 --> 01:04:08,120 So how how, how do you do that? 1060 01:04:08,120 --> 01:04:09,120 So precisely. 1061 01:04:09,120 --> 01:04:12,120 So it is every single 1062 01:04:12,320 --> 01:04:15,960 node of your functional map have a have 1063 01:04:15,960 --> 01:04:18,960 or have not got to a connection through fibers? 1064 01:04:19,000 --> 01:04:19,280 No, no. 1065 01:04:19,280 --> 01:04:22,280 The connections are only the existing connections. 1066 01:04:22,400 --> 01:04:26,360 So therefore I take the the DTI 1067 01:04:26,440 --> 01:04:27,040 with all the 1068 01:04:27,040 --> 01:04:30,840 problems that I have, but I take these as the ground through for me. 1069 01:04:30,920 --> 01:04:33,200 Okay, this is my template. 1070 01:04:33,200 --> 01:04:38,720 The good news, because in in the cases where I used the defective connectivity, 1071 01:04:38,800 --> 01:04:42,920 I only take the DTI as a mask, 1072 01:04:43,000 --> 01:04:47,600 so I will only update the connection that exists. 1073 01:04:47,680 --> 01:04:50,920 But the strain of that connection 1074 01:04:51,000 --> 01:04:54,000 does not need to correlate with the number of fibers. 1075 01:04:54,000 --> 01:04:55,840 Okay? And that is an advantage. 1076 01:04:55,840 --> 01:04:56,720 And that therefore 1077 01:04:56,720 --> 01:05:00,920 I give the name effective because I don't care about the anatomy. 1078 01:05:00,960 --> 01:05:05,240 Axes are not axes, but I decide this strength 1079 01:05:05,400 --> 01:05:09,840 based on the data, of course, but based on the functional data. 1080 01:05:09,920 --> 01:05:13,880 And they get much more than the If you try to explain functional connectivity 1081 01:05:13,960 --> 01:05:16,800 with the structure of connectivity, use, really interesting 1082 01:05:16,800 --> 01:05:20,480 say you get it'll depend on the Barcelona and blah blah blah blah. 1083 01:05:20,480 --> 01:05:26,360 But then for the standard facilitation, let's say the hundred, you get 0.3 1084 01:05:26,440 --> 01:05:28,880 a correlation, a model. 1085 01:05:28,880 --> 01:05:31,880 So for example, this tool and the model in dressing state 1086 01:05:32,000 --> 01:05:35,000 could go to 0.8 1087 01:05:35,040 --> 01:05:37,160 so that means that you are explaining much 1088 01:05:37,160 --> 01:05:40,160 more than the structure, structural connectivity. 1089 01:05:40,200 --> 01:05:42,560 And this is also known because 1090 01:05:42,560 --> 01:05:45,640 that is a very actually very simple 1091 01:05:45,680 --> 01:05:49,520 manipulation was done by the defend 1092 01:05:49,600 --> 01:05:51,680 with monkeys and by and so Douglas 1093 01:05:51,680 --> 01:05:56,160 look I love with humans one with Anastasia the other with the 1094 01:05:56,240 --> 01:05:59,400 with the sleep if you compare just the correlation 1095 01:05:59,400 --> 01:06:04,440 that I mentioned before see with FC you said 2.3, 1096 01:06:04,520 --> 01:06:09,400 but if you check that in a sleep or in anesthesia 1097 01:06:09,480 --> 01:06:10,520 goes up. 1098 01:06:10,520 --> 01:06:11,880 All right. 1099 01:06:11,880 --> 01:06:14,880 So I don't remember the number, I would say instead of point four, 1100 01:06:15,000 --> 01:06:19,200 but significantly, both at 1101 01:06:19,280 --> 01:06:24,160 and that is also entity and that the means of course the condition 1102 01:06:24,240 --> 01:06:27,920 it's not I always say 1103 01:06:28,000 --> 01:06:29,720 it's beautiful sentence 1104 01:06:29,720 --> 01:06:34,640 from from Aristotle's that was translated Latin by queen 1105 01:06:34,640 --> 01:06:38,800 and quickly through a GP to that models which is the GP do 1106 01:06:38,880 --> 01:06:42,400 which means the container shape the content 1107 01:06:42,480 --> 01:06:46,320 and say yes shape but is not determine if they can't. 1108 01:06:46,400 --> 01:06:47,000 Right. 1109 01:06:47,000 --> 01:06:51,920 So the see is really shaping FC but is not defining. 1110 01:06:52,160 --> 01:06:56,240 So the dynamic is defined in the FC 1111 01:06:56,320 --> 01:07:00,040 and therefore is not astonishing that neither sleep and and in non the sleep 1112 01:07:00,080 --> 01:07:01,560 you have different values for me 1113 01:07:01,560 --> 01:07:04,320 as demonstration that you see dynamics matters. 1114 01:07:04,320 --> 01:07:05,640 Yeah. 1115 01:07:05,640 --> 01:07:08,720 So the more the more active your brain is the less it's the less 1116 01:07:08,920 --> 01:07:12,800 the less it, the more independent Are you from this. 1117 01:07:12,880 --> 01:07:14,960 That is also a very naive 1118 01:07:14,960 --> 01:07:18,680 view of that as a come on the anatomy is pretty fixed. 1119 01:07:18,760 --> 01:07:19,200 Okay. 1120 01:07:19,200 --> 01:07:22,760 We know that's changed a little bit, but but it's pretty fixed. 1121 01:07:22,840 --> 01:07:24,600 And we are doing 1122 01:07:24,680 --> 01:07:25,480 in every 1123 01:07:25,480 --> 01:07:29,280 second of our life totally different things. 1124 01:07:29,360 --> 01:07:31,240 So how how can be 1125 01:07:31,240 --> 01:07:36,560 how could we be so flexible with the fixed, 1126 01:07:36,640 --> 01:07:41,960 boring anatomy? 1127 01:07:42,040 --> 01:07:43,520 Okay, well, if they. 1128 01:07:43,520 --> 01:07:46,760 No more questions, just think, Gustavo. 1129 01:07:46,760 --> 01:07:48,960 Once again, it is fascinating. Thank you. 1130 01:07:48,960 --> 01:07:50,120 Pleasure. Thank you.