WEBVTT 1 00:00:03.000 --> 00:00:03.930 STFC-RAL-CR03 R61: Hi! 2 00:00:04.070 --> 00:00:07.259 STFC-RAL-CR03 R61: There is a lot of documentation also. 3 00:00:07.610 --> 00:00:36.459 STFC-RAL-CR03 R61: Can we say it's been great being here? It's a very nice weather coming from Valencia, but the Uk is really cold enough. Yeah. So before I start my presentation, maybe speak a quick question. How many of you are physicists? Could you raise your hands. Okay, that makes my first.st It's like really easy to introduce. But I always want to start here, because, even though the talk is part of computing. 4 00:00:36.520 --> 00:00:58.019 STFC-RAL-CR03 R61: it's focused on applications for higher physics. So I feel it's always nice to start here. So we all know the standard model in particular physics. It's a really precise theory that describes most of the things we see in the universe, but not quite like a lot of notable examples that are missing in the standard models, such as the idea of continuum masses 5 00:00:58.020 --> 00:01:07.300 STFC-RAL-CR03 R61: that matter. So it's very clear that we need something else which we usually refer to as we understand our model physics. And this is what we are. 6 00:01:07.430 --> 00:01:17.610 STFC-RAL-CR03 R61: Me and people here around are trying to to accomplish is to find things beyond the standard model and explain that in some other way. 7 00:01:19.030 --> 00:01:42.187 STFC-RAL-CR03 R61: And also because I'm part of the collaboration, I will be focusing on this adapter, even though for the purpose of this talk, and for the purpose of application, some kind of quantum computing, it wouldn't really matter. I may as well be talking about any other experiment like. 8 00:01:42.710 --> 00:02:01.799 STFC-RAL-CR03 R61: But okay, apart from the of the collaboration, which is one of the 4 main experiments at, and we focus on the study of beauty and sandbox. But now, physics I was I was telling you about 9 00:02:02.600 --> 00:02:11.689 STFC-RAL-CR03 R61: now. If you have never seen how an looks like your back because I have one. Here, see what you see here is 10 00:02:12.030 --> 00:02:20.940 STFC-RAL-CR03 R61: the overlaid electricity vector on the actual electric panel. And then you see here the interaction point. This is where the problems connect with each other. 11 00:02:21.230 --> 00:02:24.619 STFC-RAL-CR03 R61: I'm very high now and then 12 00:02:24.810 --> 00:02:37.580 STFC-RAL-CR03 R61: with relative speeds, and then you have a bunch of things happening. So when we split those, there's a lot of particles that are produced, and they later decay at different stages of the detectors. 13 00:02:37.870 --> 00:02:52.584 STFC-RAL-CR03 R61: So of course, because it's forward so you have the condition point here, and then all the tracks and everything, because it's happening here, and the different layers of the detector. 14 00:02:53.280 --> 00:02:56.600 STFC-RAL-CR03 R61: And you can see that this is a very complicated picture. 15 00:02:56.800 --> 00:03:03.319 STFC-RAL-CR03 R61: especially if you take into consideration that you have to make out a consistent picture out of this. 16 00:03:03.350 --> 00:03:27.309 STFC-RAL-CR03 R61: identify what happened, where and how, and you need to do it at a very high rate they are working now with the bandwidth of 14 million times per second, that we have to represent an event like this one. So you can imagine that doing this efficiently and a very at a very high rate becomes really a very challenging task. From the computational point of view. 17 00:03:27.970 --> 00:03:32.639 STFC-RAL-CR03 R61: a lot of my fellow, let's see, experiments would understand when I say that 18 00:03:32.790 --> 00:03:42.087 STFC-RAL-CR03 R61: it would only get worse from the computational side, because, as time goes by, we are now in the middle of the one free week that they continued. 19 00:03:42.710 --> 00:03:46.740 STFC-RAL-CR03 R61: and then we are looking for upgrades of the experiments, which means that 20 00:03:47.030 --> 00:04:04.809 STFC-RAL-CR03 R61: higher energies will be reached and we will have more data recorded. You can increase luminosity. That increases really a lot which is very good for physics, and really good if we find if we want to find, we understand about the signatures, but it also means 21 00:04:05.380 --> 00:04:07.600 STFC-RAL-CR03 R61: that we have a talent ahead of us. 22 00:04:07.900 --> 00:04:13.060 STFC-RAL-CR03 R61: And, for example, if I'm talking from the to the side, we have very 23 00:04:13.370 --> 00:04:28.349 STFC-RAL-CR03 R61: fine tune, fancy algorithm, and it works really well. But since it's on, it's going to increase so much. 24 00:04:29.010 --> 00:04:32.950 STFC-RAL-CR03 R61: it's not going to be sustainable. So we need something else. 25 00:04:33.680 --> 00:04:40.969 STFC-RAL-CR03 R61: And this is where quantum computing comes in. So I'm going to show you a couple of slides 26 00:04:41.110 --> 00:04:54.809 STFC-RAL-CR03 R61: on quantum computing. I don't want to get into a lot of details, but I do want to explain you some of the terms because they are going to be important when we talk about pros and cons of different algorithms and also scalability, which is the main 27 00:04:55.400 --> 00:04:59.389 STFC-RAL-CR03 R61: concern about applications and computing in higher education. 28 00:05:00.310 --> 00:05:12.149 STFC-RAL-CR03 R61: So one way to look into the main difference with respect to classical computing is that we use quantum systems instead of bits. So if a bit can take 0 or one, and it's always 29 00:05:12.380 --> 00:05:16.360 STFC-RAL-CR03 R61: what it is is what you mentioned again, but it works with that 30 00:05:16.480 --> 00:05:23.260 STFC-RAL-CR03 R61: the amount of systems which means that they can benefit from fund properties like subscription. 31 00:05:23.410 --> 00:05:29.659 STFC-RAL-CR03 R61: and you know, interference. And so on. Entanglement plays a major role, which means that 32 00:05:30.000 --> 00:05:44.939 STFC-RAL-CR03 R61: you need to measure them several times because it's not like that. It's 0. It means it was a 0 at the beginning. This is not quite that simple because it's a quantum system. So you are actually collapsing by functions which means that in order to get 33 00:05:45.710 --> 00:05:53.386 STFC-RAL-CR03 R61: and accurate answer from your algorithm. You have to measure it a bunch of times because it's a probabilistic results. 34 00:05:53.870 --> 00:06:12.419 STFC-RAL-CR03 R61: And you may wonder, okay, this is doable. But it actually will play a major role, especially when we are discussing the issue of output retrieval in computer. Because one thing that I'm going to kind of sweep under the rack for the whole talk, and we'll only address at the end 35 00:06:13.000 --> 00:06:21.742 STFC-RAL-CR03 R61: is output. Retriever means that you may have the rest. Quantum algorithm. There is this quantum algorithm may have an exponential advantage. 36 00:06:22.310 --> 00:06:24.810 STFC-RAL-CR03 R61: If, when you have to measure 37 00:06:25.060 --> 00:06:36.169 STFC-RAL-CR03 R61: the output of the, you have to do it so many times that the kid of all your content. There is basically no point in having a quantum algorithm. It's really an important issue. 38 00:06:36.550 --> 00:06:52.109 STFC-RAL-CR03 R61: The reason why I will not be discussing that plan during my presentation is because it's not an issue that's specific to primary physics. It's well known issue within the community, and there are several ways to go about it. But it's less applications, more computation. 39 00:06:52.880 --> 00:07:07.050 STFC-RAL-CR03 R61: Okay, so you have your periods. And then because they are function systems that can be seen as represented in what we call the locusphere. So you have a wave function here, and then you have the eigenvalues of the same axis. 40 00:07:07.590 --> 00:07:16.269 STFC-RAL-CR03 R61: You can apply immediately. We call we call banking logic. 41 00:07:16.950 --> 00:07:21.599 STFC-RAL-CR03 R61: So you have gates and curates, and with that we can mute the content. 42 00:07:22.070 --> 00:07:29.549 STFC-RAL-CR03 R61: And then, as I've said, the main, the main reason why so many people are trying to leverage quantum computing is because 43 00:07:30.150 --> 00:07:34.259 STFC-RAL-CR03 R61: with all these properties that inherited quantum mechanics, because 44 00:07:34.570 --> 00:07:58.199 STFC-RAL-CR03 R61: dropping from superposition and I live here. To this page where where there's a very comprehensive list of algorithms in with the computational complexity. And then for some cases, there's exponential advantage product advantage. You know, there's a lot of potential in this area 45 00:08:00.366 --> 00:08:07.180 STFC-RAL-CR03 R61: about the hardware. This would be a topic for at least one second. But 46 00:08:07.430 --> 00:08:14.630 STFC-RAL-CR03 R61: one thing I want to emphasize is, there are many poor families on hybrid, and they depend they 47 00:08:14.840 --> 00:08:19.710 STFC-RAL-CR03 R61: and it benefited. From how you achieve this quantum systems. You have 48 00:08:20.290 --> 00:08:26.510 STFC-RAL-CR03 R61: new superconducting quantum computers. So Oxford quantum sequences. In this language also Ibm, who are 49 00:08:27.788 --> 00:08:35.889 STFC-RAL-CR03 R61: with Mq. Which is also is originally new events. Then how did I? 50 00:08:36.010 --> 00:08:44.050 STFC-RAL-CR03 R61: And finally, lastly, we have a special family of quantum computers, which are quantum emulators. Now. 51 00:08:45.360 --> 00:08:52.279 STFC-RAL-CR03 R61: these 3 families of quantum computers work the way I described you have units. You apply gates. 52 00:08:52.540 --> 00:08:56.259 STFC-RAL-CR03 R61: you end up with a participant, and you have your final data. 53 00:08:56.640 --> 00:08:58.890 STFC-RAL-CR03 R61: This is what again called paypaks. 54 00:08:59.830 --> 00:09:04.239 STFC-RAL-CR03 R61: This is an ecosystem in the sense that it's not case. You have your kids. 55 00:09:04.680 --> 00:09:07.539 STFC-RAL-CR03 R61: And basically they do one thing, which is 56 00:09:08.100 --> 00:09:11.799 STFC-RAL-CR03 R61: one very big thing which is minimized. And I feel like coming up 57 00:09:12.570 --> 00:09:19.029 STFC-RAL-CR03 R61: so quantum managers work very well for work, very well for minimization of isolations. 58 00:09:19.310 --> 00:09:25.310 STFC-RAL-CR03 R61: and they are especially in that sense they scale better than the other, but then they have other. 59 00:09:25.620 --> 00:09:28.880 STFC-RAL-CR03 R61: probably, or 20 family. You can think of families. 60 00:09:29.490 --> 00:09:33.150 STFC-RAL-CR03 R61: 3 advantages, and at least 3 disadvantages is a bit of 61 00:09:33.750 --> 00:09:37.406 STFC-RAL-CR03 R61: yeah, you know, it's a bit of a long term discussion of which one is going to be 62 00:09:37.920 --> 00:09:47.329 STFC-RAL-CR03 R61: to be best, and I'm going to try to keep it as agnostic as possible. So I hope you cannot tell. At the end of the seminar responsibility is my favorite. 63 00:09:47.710 --> 00:09:51.990 STFC-RAL-CR03 R61: But I will be focusing on game based program of time. So please. 64 00:09:55.060 --> 00:10:11.240 STFC-RAL-CR03 R61: now, quantum computers have been really experiencing a revolution they have evolved. You may have seen it in the news. I am very proud that they have now, I think, 400 qubits working, and they are trying to really ups upgrade this to a thousand qubits. 65 00:10:11.720 --> 00:10:14.439 STFC-RAL-CR03 R61: And the-the-microsoft is also a. 66 00:10:14.990 --> 00:10:26.678 STFC-RAL-CR03 R61: So the is really increasing. Now, what is the main challenge? The quantum computing developers have to deal with is the noise. 67 00:10:28.146 --> 00:10:35.869 STFC-RAL-CR03 R61: It's not difficult to understand that because you have a quantum system, it's kind of tricky to keep it quantum. 68 00:10:36.470 --> 00:10:42.590 STFC-RAL-CR03 R61: I think an absolute example is actually for conducting cubic needs to be an actual combat mistake. 69 00:10:42.710 --> 00:10:49.990 STFC-RAL-CR03 R61: which means very low temperatures, and also means there's a limited time in which it will be in superconducted assess. 70 00:10:51.280 --> 00:10:57.099 STFC-RAL-CR03 R61: So this is all you can. Example. Of course, in any other technologies there will be different sources of noise. 71 00:10:57.410 --> 00:11:01.260 STFC-RAL-CR03 R61: but means that all this noise is going to introduce networks. 72 00:11:01.500 --> 00:11:10.140 STFC-RAL-CR03 R61: Which means that when you measure the output from your algorithm and the output will look like something like this. 73 00:11:10.720 --> 00:11:18.260 STFC-RAL-CR03 R61: And then you have. And here the role would be the output from this from the from an actual point of computer. 74 00:11:18.400 --> 00:11:23.179 STFC-RAL-CR03 R61: the algorithm should be empty here, and because there is lines 75 00:11:23.280 --> 00:11:25.179 STFC-RAL-CR03 R61: 300 and a bit lower. 76 00:11:25.310 --> 00:11:27.130 STFC-RAL-CR03 R61: And then these entries have. 77 00:11:27.280 --> 00:11:30.339 STFC-RAL-CR03 R61: This screen has a few entries. 78 00:11:30.750 --> 00:11:43.490 STFC-RAL-CR03 R61: and this is because of noise. It really depends on the size of your circuit. A bigger circuit separate from my mind. And there are 2 79 00:11:43.870 --> 00:11:47.009 STFC-RAL-CR03 R61: ways to deal with noise depending on the time zone. 80 00:11:53.460 --> 00:11:54.260 STFC-RAL-CR03 R61: Okay. 81 00:12:01.170 --> 00:12:04.628 STFC-RAL-CR03 R61: somebody really didn't like noise. 82 00:12:05.580 --> 00:12:25.695 STFC-RAL-CR03 R61: Okay? So for the short term in quantum mediation, which means we know there's noise. We know the measurement we made is affected by noise. We are going to try to correct for this noise. Then in the longer term there's noise correction, which means we are actually going to try 83 00:12:26.220 --> 00:12:32.459 STFC-RAL-CR03 R61: to correct for the noise whilst the secret is being run in a public computer. Now. 84 00:12:33.160 --> 00:12:50.610 STFC-RAL-CR03 R61: this is important, because either way you go you will have to correct for this noise or mitigate this noise at the expense of more quantum gates or more computational processes. So it adds up everything is a bit of a trade-off between accuracy and and time. 85 00:12:51.830 --> 00:12:58.327 STFC-RAL-CR03 R61: And now to the after, I introduce you to the, to the world of quantum computing and handling physics. 86 00:12:58.790 --> 00:13:08.499 STFC-RAL-CR03 R61: we have a working group which is called Qc. For help. We were not very original with the naming, and the aim of this working group is exactly that looking for application. So 87 00:13:08.950 --> 00:13:12.129 STFC-RAL-CR03 R61: volume, computing or high energy physics. 88 00:13:12.460 --> 00:13:21.139 STFC-RAL-CR03 R61: Now, because I'm eligibility by training, I'm going to be focusing more on the experimental applications. We need 89 00:13:21.390 --> 00:13:32.920 STFC-RAL-CR03 R61: traffic construction, signal extraction, experiment, simulation. But actually, there's a whole other team dedicated to applications in field. 90 00:13:33.100 --> 00:13:37.897 STFC-RAL-CR03 R61: And I really encourage you have a look at the roadmap which we published a couple of years ago 91 00:13:38.580 --> 00:13:40.529 STFC-RAL-CR03 R61: with the same. Yet 92 00:13:41.020 --> 00:13:53.730 STFC-RAL-CR03 R61: back then, of course, because you have things like this. So this is an example of how we try to cluster the different use cases for the experimental side. So you have them clustered by topic. 93 00:13:54.490 --> 00:14:14.640 STFC-RAL-CR03 R61: cluster by type of algorithm that you will need. And then all the quantum algorithms that people have been using. Of course, this is not a definite list, but just to give a few remarkable examples. So, for example, if you have practical structure, this can be seen as a classification problem. 94 00:14:14.740 --> 00:14:17.390 STFC-RAL-CR03 R61: also aggregation optimization. 95 00:14:17.500 --> 00:14:21.140 STFC-RAL-CR03 R61: And then you have basically the sample construction you can use 96 00:14:22.120 --> 00:14:33.620 STFC-RAL-CR03 R61: most of these algorithms. But just to give you an idea on where the field is going, also, because really rapidly evolving groups, there is a lot of interest also from 97 00:14:33.810 --> 00:14:37.289 STFC-RAL-CR03 R61: private companies. So it's good to keep an eye on that. 98 00:14:38.830 --> 00:14:46.214 STFC-RAL-CR03 R61: Now there are many, many different topics. I cannot emphasize this enough. So obviously, I will be focusing on the ones 99 00:14:46.650 --> 00:14:48.400 STFC-RAL-CR03 R61: I work the most. 100 00:14:48.840 --> 00:15:01.779 STFC-RAL-CR03 R61: But that doesn't mean that the other topics are any less interesting. So I wanted to give you a taste on the things that I cannot cover in this talk like, for example, if you think on the theory side, that is Uc. 101 00:15:02.050 --> 00:15:09.020 STFC-RAL-CR03 R61: you can imagine that simulations in that specific could really benefit from quantum computer. And then they are 102 00:15:09.210 --> 00:15:14.450 STFC-RAL-CR03 R61: running these on simplified models. I'll give you a reference, but it is actually underneath 103 00:15:14.890 --> 00:15:21.699 STFC-RAL-CR03 R61: a very profitable way to use computing for or the high energy physics. 104 00:15:21.870 --> 00:15:41.040 STFC-RAL-CR03 R61: There's also neutrino oscillations that use quantum computing. Now, this is a bit of a special case, because neutrino oscillations, since there are 3 flavors of neutrinos they can actually profit from, not the qubits which have 2 Meg levels, but qubits which have 3 nap levels. 105 00:15:41.410 --> 00:15:50.909 STFC-RAL-CR03 R61: So it's a system which in practice will have 3 different Eigen values. And then you can encode. The flavor of your laptop. 106 00:15:51.600 --> 00:15:55.395 STFC-RAL-CR03 R61: Strategic as anything that's beyond 2 energy levels 107 00:15:56.440 --> 00:16:05.160 STFC-RAL-CR03 R61: are a bit behind on hardware for obvious reasons, because it's keeping 2 energy levels is tricky. You cannot imagine what happened in Ohio. 108 00:16:06.000 --> 00:16:13.820 STFC-RAL-CR03 R61: but there's actually been a lot of development, and they've done this very good on my cell collection for this. So since what I'm missing. 109 00:16:14.500 --> 00:16:32.109 STFC-RAL-CR03 R61: and lastly, the Permian wave package is happening which tries to use quantum mechanical evolution to do some theory calculations, so I leave you just a very terrific list of examples. But if you go to the roadmap paper you can find a bunch of options. 110 00:16:33.370 --> 00:16:41.019 STFC-RAL-CR03 R61: And now to my main topic of research. That's 5. Computing is quantum computing and traffic construction. Now. 111 00:16:41.940 --> 00:16:45.840 STFC-RAL-CR03 R61: when we think about applications of quantum computing. 112 00:16:45.950 --> 00:16:56.846 STFC-RAL-CR03 R61: the quantum advantage is actually tricky to define, because it depends on what your problem is, how optimistic you are and what are you looking at? You know some people 113 00:16:57.760 --> 00:17:05.249 STFC-RAL-CR03 R61: talk about advantage in terms of computational complexity. So if your evolution scales exponentially. 114 00:17:05.430 --> 00:17:16.339 STFC-RAL-CR03 R61: some people talk about it in terms of timing in terms, even in terms of sustainability, you could have a discussion about this. 115 00:17:16.880 --> 00:17:27.120 STFC-RAL-CR03 R61: And then in high energy physics. We've had 3 annual meetings of the physical health community, and we haven't managed to agree 116 00:17:27.400 --> 00:17:32.260 STFC-RAL-CR03 R61: on. What are we looking for when we talk about quantum advantage for high ophysics? 117 00:17:32.690 --> 00:17:52.669 STFC-RAL-CR03 R61: Because the thing is, if you're a practical physicist, you put mind processes are basically quantum in nature. We are measuring quantum processes, quantum theory. That's what plays a big role in our case. So it seems like a perfect match. It seems like something that could work perfectly. It's like the best choice. 118 00:17:53.310 --> 00:17:54.457 STFC-RAL-CR03 R61: But of course 119 00:17:55.180 --> 00:18:03.720 STFC-RAL-CR03 R61: this only works for very certain use, cases like for experiment simulation. It would work because you are simulating events. But 120 00:18:04.410 --> 00:18:13.399 STFC-RAL-CR03 R61: the moment you are taking data, this is quantum, yes, but the latest. So everything that was quantum about it. 121 00:18:13.720 --> 00:18:16.790 STFC-RAL-CR03 R61: It's kind of lost. You cannot look at it that way anymore. 122 00:18:17.070 --> 00:18:23.729 STFC-RAL-CR03 R61: Which means that you have to deal with quantum computing, not as a type, that match because you're matching phantom with phantom. 123 00:18:23.890 --> 00:18:30.389 STFC-RAL-CR03 R61: But as a computational tool. So you have to look for the algorithm that gives you the more 124 00:18:30.690 --> 00:18:41.280 STFC-RAL-CR03 R61: the highest computational requirements. And some people really don't like that, because it seems like, you know, going quantum very classical, then quantum again. 125 00:18:41.720 --> 00:18:46.660 STFC-RAL-CR03 R61: And there are ideas in which you could in principle bypass this classical part. 126 00:18:47.030 --> 00:18:56.209 STFC-RAL-CR03 R61: But since we have to talk in the context of 40 years, I don't think this will happen at the Lhc in 40 years. So I'm going to keep talking about 127 00:18:56.470 --> 00:18:58.920 STFC-RAL-CR03 R61: classical data in traffic instruction. 128 00:19:00.560 --> 00:19:05.920 STFC-RAL-CR03 R61: Now, the problem of traffic consumption is a very common problem to any client based experiment. 129 00:19:06.598 --> 00:19:15.870 STFC-RAL-CR03 R61: So I put one Atlas image here in my presentation to emphasize that it's actually kind of experiment agnostic. 130 00:19:16.130 --> 00:19:28.649 STFC-RAL-CR03 R61: The thing is, you will have collision points, and then just particles travel travels through the detector to the different layers that you have in the detector, and we leave signals. And then these signals and 131 00:19:29.080 --> 00:19:32.810 STFC-RAL-CR03 R61: 3 dimensional points which are hit. I'm not considering you 132 00:19:33.200 --> 00:19:42.720 STFC-RAL-CR03 R61: save the time. I know this is a possibility in the future, but for the moment. Let's consider special coordinates as what their fields are composed of. 133 00:19:43.300 --> 00:19:55.059 STFC-RAL-CR03 R61: And then your task is with all these kids. You have to make a consistent picture that gives you the tracks so that you can use it for everybody, everywhere else in the in the experiment. 134 00:19:56.010 --> 00:20:10.220 STFC-RAL-CR03 R61: So this problem is very interesting from the computational perspective. Very challenging, because you have a very large number of tracks, leads to a very large number of segments. Segments are a couple of hits 135 00:20:10.520 --> 00:20:19.739 STFC-RAL-CR03 R61: in consecutive layers, and also very important, because, like reconstruction is that under himmy, and it populates 136 00:20:19.960 --> 00:20:23.710 STFC-RAL-CR03 R61: everything. That's I mean, of course, if you're dealing with new contract. Of course not. 137 00:20:24.300 --> 00:20:29.180 STFC-RAL-CR03 R61: But unless you're dealing with that, it basically populates everything that you're doing with it. 138 00:20:29.650 --> 00:20:34.869 STFC-RAL-CR03 R61: So this is why a lot of people in the funding computing community try to 139 00:20:35.140 --> 00:20:43.230 STFC-RAL-CR03 R61: focus that. I focus on this because it seems like one of the biggest computational challenges that we needed to address and solve. 140 00:20:45.251 --> 00:20:50.989 STFC-RAL-CR03 R61: So in the classical world which we are living in now as 18, 141 00:20:51.210 --> 00:21:03.110 STFC-RAL-CR03 R61: we usually do it in 2 ways. This is not a black and white definition. But for the purpose of the talk we want to proceed. It's not actually that way. 142 00:21:04.023 --> 00:21:08.900 STFC-RAL-CR03 R61: But we basically have local tracking methods and global tracking methods. 143 00:21:09.580 --> 00:21:14.220 STFC-RAL-CR03 R61: So local traffic detecting methods means you take a series of steps 144 00:21:14.600 --> 00:21:19.519 STFC-RAL-CR03 R61: to go from here where you have all the hits to here where you have all the tracks. 145 00:21:20.210 --> 00:21:28.609 STFC-RAL-CR03 R61: These are steps are usually something like sitting. Then you have stepulating of the of the tablets, you know. Then you have some cleanup. 146 00:21:29.200 --> 00:21:34.200 STFC-RAL-CR03 R61: Then you have some some of processing, and you end up with the trucks. 147 00:21:34.520 --> 00:21:42.257 STFC-RAL-CR03 R61: So it's a sequential approach. And this is what Ltv does. This is what the seeing my stuff. 148 00:21:42.790 --> 00:21:46.569 STFC-RAL-CR03 R61: It is what Atlas does. So it's kind of sequential. 149 00:21:47.280 --> 00:21:54.709 STFC-RAL-CR03 R61: And then you can also consider global counting records which means ambient. You recall the hits. 150 00:21:54.920 --> 00:21:57.070 STFC-RAL-CR03 R61: you insert the library down here. 151 00:21:57.200 --> 00:21:59.110 STFC-RAL-CR03 R61: You learn about the tracks. 152 00:22:00.220 --> 00:22:07.720 STFC-RAL-CR03 R61: Of course, ideally, this is fantastic with all the hits, and then one algorithm will will give you all the draft. 153 00:22:08.650 --> 00:22:11.823 STFC-RAL-CR03 R61: Of course it will not be the case. 154 00:22:12.730 --> 00:22:15.729 STFC-RAL-CR03 R61: if it were that easy it would have been used classically. 155 00:22:16.420 --> 00:22:26.399 STFC-RAL-CR03 R61: but it makes sense that it's where we are going to focus on computing. There's actually an study done, a local method. And the the improvement was 156 00:22:26.620 --> 00:22:31.169 STFC-RAL-CR03 R61: for that to get best in one of the steps. So it was not very promising. 157 00:22:31.490 --> 00:22:38.430 STFC-RAL-CR03 R61: There was something to be learned, but it was not very promising, because the algorithms were already so well tuned to the needs of the detectors. 158 00:22:39.330 --> 00:22:44.520 STFC-RAL-CR03 R61: So we went with local, with local tracking methods. And this is what I'm going to focus on. 159 00:22:46.850 --> 00:22:53.779 STFC-RAL-CR03 R61: So the way we do this, and the when I say we is, as far as I know. 160 00:22:53.960 --> 00:23:00.929 STFC-RAL-CR03 R61: everybody that has tried to do. Quantum computing from traffic construction is kind of going about it in the same way. 161 00:23:01.490 --> 00:23:08.160 STFC-RAL-CR03 R61: So, as I've said, you can expect to have some improvements in, even in complexity, timing complexity. 162 00:23:08.660 --> 00:23:12.114 STFC-RAL-CR03 R61: So the way we try to really have the construction in quantum is 163 00:23:12.740 --> 00:23:21.010 STFC-RAL-CR03 R61: we tried to describe the driver construction program as an icing, like onion, which has a logic dependence, and a native dependence 164 00:23:21.630 --> 00:23:29.450 STFC-RAL-CR03 R61: such that the ground state of this Hamiltonian will contain the solution of the traffic construction problem. 165 00:23:30.370 --> 00:23:34.750 STFC-RAL-CR03 R61: Now, why do we all try to do this this way? 166 00:23:35.370 --> 00:23:37.334 STFC-RAL-CR03 R61: This is the person wants to contribute in 167 00:23:37.710 --> 00:23:40.930 STFC-RAL-CR03 R61: once you get to the point of having an excellent. 168 00:23:41.630 --> 00:23:51.070 STFC-RAL-CR03 R61: That's the best spot. Because if there are team, quantum chemistry, quantum biology, protein, simulation, molecular simulation model. 169 00:23:51.680 --> 00:24:01.000 STFC-RAL-CR03 R61: there are a lot of algorithms out there, but especially done to minimize this also fund managers can minimize this very well. 170 00:24:02.200 --> 00:24:12.469 STFC-RAL-CR03 R61: So the problem is actually how you formulate the Hamiltonian. And you want to do it in a way that's general enough that you can deal with all your events. 171 00:24:13.050 --> 00:24:17.915 STFC-RAL-CR03 R61: but also using some information from your detect from your detector, because otherwise 172 00:24:18.440 --> 00:24:22.860 STFC-RAL-CR03 R61: your algorithm will never minimized and you will never find solution. 173 00:24:23.340 --> 00:24:25.720 STFC-RAL-CR03 R61: So this is rapidly the approach we take. 174 00:24:27.830 --> 00:24:34.460 STFC-RAL-CR03 R61: And I'm going to give you 2 3 examples that we have tried, we, the principal within community 175 00:24:34.720 --> 00:24:40.950 STFC-RAL-CR03 R61: and to try and and deal with that construction with quantum computing. So the 1st is the digital algorithm. 176 00:24:41.190 --> 00:24:56.820 STFC-RAL-CR03 R61: Now, this algorithm is kind of a conservation algorithm in the content community because it has an exponential advantage which is really very good to classify. But it comes with a series with a series of very stringent conditions on 177 00:24:57.070 --> 00:25:01.769 STFC-RAL-CR03 R61: the use case, it will only be good to use. If 178 00:25:02.040 --> 00:25:07.619 STFC-RAL-CR03 R61: so, it's designed to solve a system of linear equations. And we observe this, it will only work 179 00:25:08.020 --> 00:25:15.720 STFC-RAL-CR03 R61: if your matrix is semi positive. That means, first, st if you can initialize this vector here 180 00:25:16.280 --> 00:25:22.059 STFC-RAL-CR03 R61: quite randomly, it doesn't need to be in a special mistake and so on. So it's really 181 00:25:22.520 --> 00:25:24.669 STFC-RAL-CR03 R61: use case for this. 182 00:25:25.820 --> 00:25:35.579 STFC-RAL-CR03 R61: the way we have tried to do this is, we have tried to define a like the one I showed you a couple of slides ago, so I see 9 183 00:25:36.640 --> 00:25:38.990 STFC-RAL-CR03 R61: in terms of the segments. Now 184 00:25:39.120 --> 00:25:43.029 STFC-RAL-CR03 R61: going to walk you through the 3 terms that we have in this environment here. 185 00:25:43.850 --> 00:26:01.919 STFC-RAL-CR03 R61: This is a Hamiltonian. That's so. These are binary values that defeat to your quantum, and then at the end 186 00:26:03.130 --> 00:26:11.423 STFC-RAL-CR03 R61: they would be 0. If the segment does not belong to a track, which means that the 2 heaps that form the segment do not belong to the same track. 187 00:26:11.830 --> 00:26:15.949 STFC-RAL-CR03 R61: and it will be one or all if they do bring up to the same plan. 188 00:26:17.090 --> 00:26:23.231 STFC-RAL-CR03 R61: Now, you have a poetic term that's basically making the consecutive segments be straight. 189 00:26:24.220 --> 00:26:30.489 STFC-RAL-CR03 R61: Why is this? Because it's already challenging enough to begin with the straight tracks. So straight tracks. 190 00:26:30.620 --> 00:26:33.852 STFC-RAL-CR03 R61: Yeah, so yeah. 191 00:26:35.390 --> 00:26:42.700 STFC-RAL-CR03 R61: And then you have 2 penalty terms. And the penalty terms are a bit. 192 00:26:43.070 --> 00:26:46.979 STFC-RAL-CR03 R61: And it really depends in this case. But they are doing it. They are trying 193 00:26:47.160 --> 00:26:49.820 STFC-RAL-CR03 R61: to make the editorial algorithm work. 194 00:26:49.920 --> 00:27:05.269 STFC-RAL-CR03 R61: So this 1st term making the spectrum of the matrix positive, because that's what. And this other term is making sure that your ground state in the algorithm is well separated from the rest. 195 00:27:05.930 --> 00:27:11.240 STFC-RAL-CR03 R61: And then you have with high per parameters that you basically 196 00:27:11.670 --> 00:27:18.260 STFC-RAL-CR03 R61: give depending on the way you have, you want this penalty terms to have. So if they are higher, it means that 197 00:27:18.770 --> 00:27:25.749 STFC-RAL-CR03 R61: they are more important. When we miss Hamiltonian. If they are lower, they- they are not that important. They- they play a small part. 198 00:27:26.768 --> 00:27:34.309 STFC-RAL-CR03 R61: So once so once you fit this. All the parameters that you see in there are fixed. 199 00:27:34.600 --> 00:27:40.389 STFC-RAL-CR03 R61: You have done a harder time themselves to try to find the optimal values and memory updates. 200 00:27:42.410 --> 00:27:49.339 STFC-RAL-CR03 R61: And we have tried this for analytics, and it works quite well. 201 00:27:49.450 --> 00:27:58.570 STFC-RAL-CR03 R61: So the red spots that you see here are the collision points from a single incident at the 1st part of the electrical, because that's what the 202 00:27:59.290 --> 00:28:06.189 STFC-RAL-CR03 R61: the Chat Transcript, and then, you see, invite the representative text. 203 00:28:06.430 --> 00:28:09.040 STFC-RAL-CR03 R61: So where you see it's not perfect. 204 00:28:09.840 --> 00:28:21.319 STFC-RAL-CR03 R61: It works quite well, and they usually performance metrics that we work with, such as reconstruction, efficiency, or he purely a kind of 205 00:28:21.620 --> 00:28:27.000 STFC-RAL-CR03 R61: home file which is safe at the end with a lot of oops. 206 00:28:28.470 --> 00:28:29.360 STFC-RAL-CR03 R61: No. 207 00:28:30.400 --> 00:28:47.799 STFC-RAL-CR03 R61: it works very well. We were very down to the result. It was a very important milestone, but of course there's a trick, and that is that it's been validated on the classic and linear solver, which means that we used the Hamiltonian that I showed you. 208 00:28:48.010 --> 00:28:52.810 STFC-RAL-CR03 R61: The algorithm is analog to the to the ATM, but it's classically. 209 00:28:53.430 --> 00:28:54.730 STFC-RAL-CR03 R61: And where is this? 210 00:28:55.130 --> 00:29:01.730 STFC-RAL-CR03 R61: Because the editor has a problem that SQL. Was far, far too big to run. 211 00:29:02.950 --> 00:29:12.730 STFC-RAL-CR03 R61: 4 tracks and 60 dental layers took 2 days in a quantum simulator, so you can only imagine what a thousand tracks and 50 dental layers will do. 212 00:29:13.550 --> 00:29:23.149 STFC-RAL-CR03 R61: And don't even think about running this on an actual quantum computer, because the system is just too too many gates, and we cannot feel it in a reasonable quantum computer. 213 00:29:23.990 --> 00:29:37.378 STFC-RAL-CR03 R61: So after we kind of got this result that this was working well in simulation, we started to focus on how to improve this circuit, to make it smaller. And this is where the funding computing 214 00:29:38.050 --> 00:29:43.239 STFC-RAL-CR03 R61: computational aspect comes into play. Because basically, what you do is you say, okay. 215 00:29:44.070 --> 00:29:56.550 STFC-RAL-CR03 R61: for example, how is built? How is the algorithm actually built and turns out there? One part of the partnership you know a little about quantum computing is quantum based estimation 216 00:29:56.730 --> 00:30:00.619 STFC-RAL-CR03 R61: which we we were taking as a default. But actually, we really need 217 00:30:00.720 --> 00:30:03.640 STFC-RAL-CR03 R61: so much position. So we try to reduce that. 218 00:30:04.110 --> 00:30:10.529 STFC-RAL-CR03 R61: And so the results were very promising in that you can see if it was very reduced. But it's it's still 219 00:30:11.920 --> 00:30:17.480 STFC-RAL-CR03 R61: in the very long term, and Ibm lives up to the promise that we can manage. 220 00:30:18.890 --> 00:30:23.770 STFC-RAL-CR03 R61: How much do you know that you've done? You've done cool. 221 00:30:24.220 --> 00:30:30.249 STFC-RAL-CR03 R61: Okay? So this was one algorithm, we can choose, we can choose from a really 1 million of them. 222 00:30:31.177 --> 00:30:38.189 STFC-RAL-CR03 R61: So next we tried with Q, anyway. Now, this algorithm stands for quantum approximate optimization. 223 00:30:38.700 --> 00:30:53.429 STFC-RAL-CR03 R61: This is one of the sort of canonical algorithm that works as a vacational algorithm. And the main task it has is solving communal problems like the traffic mustache. If I told you all the heats 224 00:30:53.630 --> 00:31:04.339 STFC-RAL-CR03 R61: I can't tell you is anything related, you can think of it as can also be regarded as this type of program. But you can think about it 225 00:31:04.470 --> 00:31:16.170 STFC-RAL-CR03 R61: in terms of graph theory. So it's a very one type of problem. And it has this algorithm which theoretically assures you that you can find 226 00:31:16.360 --> 00:31:21.090 STFC-RAL-CR03 R61: the ground state within reason. Approximate means. 227 00:31:21.510 --> 00:31:26.350 STFC-RAL-CR03 R61: It will never to be like 100%, because this is an empty heart problem. 228 00:31:26.690 --> 00:31:30.089 STFC-RAL-CR03 R61: so it cannot, but it will find it a personal number. 229 00:31:31.070 --> 00:31:33.350 STFC-RAL-CR03 R61: Now, the way it works is 230 00:31:33.600 --> 00:31:37.990 STFC-RAL-CR03 R61: it's like an approximation of what the quantum engineering does. So you have 231 00:31:38.370 --> 00:31:42.109 STFC-RAL-CR03 R61: your problem, which is the one that you're trying to minimize. 232 00:31:42.780 --> 00:31:49.819 STFC-RAL-CR03 R61: I'll give you the unitary matrix with the learner scales and a 1 parameter problem. 233 00:31:50.530 --> 00:31:55.999 STFC-RAL-CR03 R61: and then you have a mix in 100,000 HP. With the hybrid parameters. 234 00:31:56.570 --> 00:32:01.200 STFC-RAL-CR03 R61: Now, this needs in California as you want it. 235 00:32:01.390 --> 00:32:09.480 STFC-RAL-CR03 R61: So this is how it was a series of what it's doing is it's making sure that your system 236 00:32:09.600 --> 00:32:21.509 STFC-RAL-CR03 R61: it will not minimize. So you need it. You don't need it to be very complex, but you need it. 237 00:32:22.543 --> 00:32:26.550 STFC-RAL-CR03 R61: So this is an intense evaluation. You prepare your phone to mistake 238 00:32:27.500 --> 00:32:34.540 STFC-RAL-CR03 R61: in a way that looks like if you applied. Is this limit? There is a number of times that are the layers of the algorithm. 239 00:32:36.280 --> 00:32:43.410 STFC-RAL-CR03 R61: Then you compute the expectation values of the energy of your problem hydrogen. So the energy of your state. 240 00:32:43.540 --> 00:32:48.269 STFC-RAL-CR03 R61: And then you optimize this hyperparameters, Beta and Xaml classically. 241 00:32:48.590 --> 00:32:56.725 STFC-RAL-CR03 R61: So, a classical optimizer will choose the best values and then iterate until you find they weren't safe again. 242 00:32:58.140 --> 00:33:03.559 STFC-RAL-CR03 R61: So in this case, as you can see, you actually have 2 hyper. 243 00:33:04.550 --> 00:33:09.750 STFC-RAL-CR03 R61: And now why did we try to go for this because 244 00:33:10.266 --> 00:33:17.399 STFC-RAL-CR03 R61: I was telling you how edited produces really big secrets. You have one complete system management. It's a complicated secret. 245 00:33:18.286 --> 00:33:22.850 STFC-RAL-CR03 R61: Which means that in the in the near term devices 246 00:33:22.990 --> 00:33:25.809 STFC-RAL-CR03 R61: we cannot run it for sure, no matter 247 00:33:26.000 --> 00:33:29.549 STFC-RAL-CR03 R61: how many fancy techniques we try to to use to reduce the size 248 00:33:30.250 --> 00:33:41.989 STFC-RAL-CR03 R61: for Q. And A. It's actually thought to be better for near term devices. It's considered more resistant to noise, and it's supposedly better at producing smaller cities. 249 00:33:42.190 --> 00:33:45.259 STFC-RAL-CR03 R61: I say, supposedly, because for this case it really was not 250 00:33:46.870 --> 00:33:50.090 STFC-RAL-CR03 R61: so. In this case we use a slightly different comment. 251 00:33:50.520 --> 00:33:56.310 STFC-RAL-CR03 R61: It looks very different. It's actually probably different here this time is doing the same as before is 252 00:33:56.490 --> 00:34:03.040 STFC-RAL-CR03 R61: different, the segments of, and then I have 2 terms here in this case 253 00:34:07.230 --> 00:34:12.880 STFC-RAL-CR03 R61: faster to decide the nanometers by keeping the metrics as far as or whatever. 254 00:34:13.040 --> 00:34:22.589 STFC-RAL-CR03 R61: What we need is, we impose that segment having consecutive layers. So this that is, for bringing this 255 00:34:22.889 --> 00:34:27.920 STFC-RAL-CR03 R61: to belong to the same time, you know, because if they share heads of deals. 256 00:34:28.110 --> 00:34:29.960 STFC-RAL-CR03 R61: this is uncritical for a track. 257 00:34:30.659 --> 00:34:36.790 STFC-RAL-CR03 R61: And this time is basically say it's an organization that it's selling the other event. 258 00:34:37.780 --> 00:34:41.200 STFC-RAL-CR03 R61: What is the number of active segments? It should find? 259 00:34:41.489 --> 00:34:45.930 STFC-RAL-CR03 R61: The reason why this term is done is because if you remove yourself 260 00:34:46.199 --> 00:34:54.119 STFC-RAL-CR03 R61: and you need your algorithm to converge, your algorithm does not care about your physics does not care about the rank of retrieval. So 261 00:34:54.409 --> 00:34:58.959 STFC-RAL-CR03 R61: if I remove this, my algorithm will basically say. 262 00:34:59.810 --> 00:35:02.959 STFC-RAL-CR03 R61: all segments are correct and solution is in there. 263 00:35:03.650 --> 00:35:10.239 STFC-RAL-CR03 R61: It's saying that it's true. But it's not like what? So you have. We have to to use the solution like that 264 00:35:12.450 --> 00:35:21.160 STFC-RAL-CR03 R61: now. It was a very promising idea. But we quickly realize that it has really big discovery issues. 265 00:35:21.932 --> 00:35:23.750 STFC-RAL-CR03 R61: So you want to complete it. 266 00:35:24.080 --> 00:35:31.649 STFC-RAL-CR03 R61: When you have a quantum system, you can parameterize this as a molecules and a phase which means that you can encode the information phase. 267 00:35:32.050 --> 00:35:37.880 STFC-RAL-CR03 R61: or, you know, in amplitude, there are some ways to go about which means that 268 00:35:38.330 --> 00:35:48.680 STFC-RAL-CR03 R61: you do have a problem of having to use classical data. But the problem is actually reduced because of the same code. So, for example, for 8 itself, the number of qubits was not the issue. We could 269 00:35:48.800 --> 00:35:50.899 STFC-RAL-CR03 R61: book a lot of traps with 270 00:35:51.240 --> 00:35:59.620 STFC-RAL-CR03 R61: a realistic amount of queues. So this was good with Qa. Is really not the case. We cannot do any sort of thing, because 271 00:36:00.230 --> 00:36:05.770 STFC-RAL-CR03 R61: the Hamiltonian is is list, says that Portugal segment students. 272 00:36:06.200 --> 00:36:13.510 STFC-RAL-CR03 R61: which means that you have 2 tracks, 3 layers, literally the simplest example you can think of in periods. 273 00:36:14.970 --> 00:36:25.220 STFC-RAL-CR03 R61: and it only gets worse because it's a question of all possible segments, right? Unless you do whatever it only grows, and it grows very fast. 274 00:36:25.640 --> 00:36:31.259 STFC-RAL-CR03 R61: So, for example, when we reach T tracks 4 layers, the simulator gave up. 275 00:36:32.100 --> 00:36:46.040 STFC-RAL-CR03 R61: and the reason why this is the case in the simulator is because you are asking simulator a classical computer to simulate 27 quantum systems and amplitudes. You dangle them. 276 00:36:46.210 --> 00:36:51.189 STFC-RAL-CR03 R61: And we made a lot of operations of that we made like a thousand operations 277 00:36:53.000 --> 00:36:56.380 STFC-RAL-CR03 R61: now. Yes, they were very simple operations. But they were operations. 278 00:36:56.920 --> 00:37:02.850 STFC-RAL-CR03 R61: So basically, the classical computer gave up because of memory issues. Essentially. 279 00:37:03.150 --> 00:37:09.179 STFC-RAL-CR03 R61: ironically enough, if you were to run this on an actual computer, it could actually run better. 280 00:37:09.890 --> 00:37:18.650 STFC-RAL-CR03 R61: Now, this seems like a somewhat irreducible issues issue in that either. Routines account. 281 00:37:18.870 --> 00:37:25.630 STFC-RAL-CR03 R61: We have tried to do a number of things we have tried because this algorithm is all going wrong and so badly used. 282 00:37:25.750 --> 00:37:28.570 STFC-RAL-CR03 R61: There are many ways to try to improve it 283 00:37:28.690 --> 00:37:34.769 STFC-RAL-CR03 R61: which are usually called modified queue, anyway. So we're actually focusing 284 00:37:37.030 --> 00:37:39.680 STFC-RAL-CR03 R61: for reducing the number of layers here. 285 00:37:39.800 --> 00:37:42.569 STFC-RAL-CR03 R61: So this is one layer. This is another layer. 286 00:37:42.690 --> 00:37:44.390 STFC-RAL-CR03 R61: The more layers you have. 287 00:37:44.930 --> 00:37:47.620 STFC-RAL-CR03 R61: Maybe if you're sick with this, and 288 00:37:48.729 --> 00:37:54.670 STFC-RAL-CR03 R61: from some point from 12 qubits, you need at least 2 languages for the algorithm to convert. 289 00:37:55.650 --> 00:38:02.620 STFC-RAL-CR03 R61: So this you can do, it could help with the size of your circuit. It will definitely not solve this kind of issue 290 00:38:03.430 --> 00:38:07.969 STFC-RAL-CR03 R61: and goodbye triplets instead of doublets because 291 00:38:08.300 --> 00:38:16.820 STFC-RAL-CR03 R61: possible segments. Are 2 hits. But when actually take 3 hits was actually close to 292 00:38:17.280 --> 00:38:20.040 STFC-RAL-CR03 R61: in worsening scalability by a lot. 293 00:38:20.410 --> 00:38:24.230 STFC-RAL-CR03 R61: There are some other groups who have tried to rewrite this problem. And I say. 294 00:38:24.560 --> 00:38:30.950 STFC-RAL-CR03 R61: Okay, I only take 5 tracks at a time, because that's all my volume can handle, and then 295 00:38:31.100 --> 00:38:37.839 STFC-RAL-CR03 R61: don't go ahead. Sorry. Okay. 296 00:38:38.160 --> 00:38:46.420 STFC-RAL-CR03 R61: this is this is what I did. But I think we've reached this point, and this is a conclusion. It really does not escape. Well, we define different approach. 297 00:38:46.630 --> 00:38:49.830 STFC-RAL-CR03 R61: or you know, professional. It's not 298 00:38:50.520 --> 00:38:53.260 STFC-RAL-CR03 R61: meant to rely on actual part of the webinar. 299 00:38:54.780 --> 00:38:59.510 STFC-RAL-CR03 R61: No, they are very simple for this type of approach. 300 00:39:00.221 --> 00:39:05.470 STFC-RAL-CR03 R61: Apart from what another group we basically did, which I want to show you in the next slide. 301 00:39:05.680 --> 00:39:11.850 STFC-RAL-CR03 R61: There's a way to kind of try to improve this type of algorithms, because. 302 00:39:12.050 --> 00:39:19.050 STFC-RAL-CR03 R61: you know, stop family that's called variation of quantum algorithms, which means that you feed the circuit instead of 303 00:39:19.250 --> 00:39:22.639 STFC-RAL-CR03 R61: parameters that you buy until you find an optimal 304 00:39:23.070 --> 00:39:25.329 STFC-RAL-CR03 R61: solution for what you want it. 305 00:39:26.050 --> 00:39:34.430 STFC-RAL-CR03 R61: They are. They are usually hybrid classical quantum, so the variation part can go in. The classical interior can go in the quantum. 306 00:39:35.190 --> 00:39:44.479 STFC-RAL-CR03 R61: The problem with with this quantum algorithms is that we have one very important issue which has bandwidth. 307 00:39:46.850 --> 00:39:51.099 STFC-RAL-CR03 R61: But even if we could run this QA. And it would run. 308 00:39:51.380 --> 00:39:53.479 STFC-RAL-CR03 R61: Juliets were not not an issue. 309 00:39:54.663 --> 00:39:59.759 STFC-RAL-CR03 R61: There are places in the face space where it gets stuck. 310 00:39:59.990 --> 00:40:00.890 STFC-RAL-CR03 R61: Here I go. 311 00:40:01.330 --> 00:40:08.117 STFC-RAL-CR03 R61: The diversion gradient vanishes so it- it stops minimizing. It just gets stuck in there. 312 00:40:08.850 --> 00:40:10.899 STFC-RAL-CR03 R61: And he was not realizing it. But 313 00:40:11.360 --> 00:40:16.460 STFC-RAL-CR03 R61: and the the problem about this type of problems for a variational algorithm 314 00:40:16.800 --> 00:40:25.760 STFC-RAL-CR03 R61: like is that they are not easy to see, because if you have a Hamiltonian, if you see the iterations, you can see. 315 00:40:25.940 --> 00:40:30.871 STFC-RAL-CR03 R61: you can try to understand. If you understand, in a local email or the Admin portal. 316 00:40:31.510 --> 00:40:36.589 STFC-RAL-CR03 R61: but they are not very easy to see, and if you can see them they are not easy to avoid. 317 00:40:37.610 --> 00:40:40.491 STFC-RAL-CR03 R61: So there's a lot of work in the communication community. 318 00:40:41.000 --> 00:40:48.090 STFC-RAL-CR03 R61: And because we are a physicist, we're trying to observe this from a physics perspective. So what we we 319 00:40:48.440 --> 00:40:56.016 STFC-RAL-CR03 R61: in this case I'm not included people from this here, and also from my- my home, my home institution. In mind you have, that is. 320 00:40:56.710 --> 00:41:01.830 STFC-RAL-CR03 R61: instead of using Iranian design. They try to use the in 2 months. 321 00:41:02.800 --> 00:41:08.380 STFC-RAL-CR03 R61: and this is the only accurate sentence I can give you about this, because it's a fairly complicated allocation 322 00:41:08.720 --> 00:41:15.436 STFC-RAL-CR03 R61: webinar, what they claim in this paper here, and also in the presentation that they- they gave an 323 00:41:17.069 --> 00:41:22.789 STFC-RAL-CR03 R61: the libraries, in the sense, is to blame because there wasn't cut off. 324 00:41:23.000 --> 00:41:26.509 STFC-RAL-CR03 R61: And if you could use something which you see 325 00:41:27.030 --> 00:41:31.614 STFC-RAL-CR03 R61: inheritance which you can define in theory. 326 00:41:32.600 --> 00:41:46.850 STFC-RAL-CR03 R61: then we put about 500,000, because I find this very highly interesting. I did read this paper because they basically concluded that they were better than which they are 327 00:41:47.469 --> 00:41:52.920 STFC-RAL-CR03 R61: but again, scalability. So I couldn't implement it. But I hope somebody can. 328 00:41:54.240 --> 00:42:11.510 STFC-RAL-CR03 R61: And now, not to make my whole presentation. Ledc, these are my 2 slides that are not. They actually have 329 00:42:11.640 --> 00:42:18.461 STFC-RAL-CR03 R61: plans for this experiment, which is repositor groups. And here we have 4 layers in the tracking system. 330 00:42:19.060 --> 00:42:25.259 STFC-RAL-CR03 R61: So it actually, you know, fits what you know it's 4 layers instead of 50. So. 331 00:42:25.770 --> 00:42:35.299 STFC-RAL-CR03 R61: and what the basic people tried is the same approach and isolate from your domain with one difference, that is, they consider triplets instead of segments. 332 00:42:35.840 --> 00:42:42.879 STFC-RAL-CR03 R61: and they try 3 methods. They try variational algorithms which were highly quantum classic. 333 00:42:43.070 --> 00:42:45.930 STFC-RAL-CR03 R61: They find a standard communal retirement filter. 334 00:42:46.310 --> 00:42:51.459 STFC-RAL-CR03 R61: and they try graph neural networks, both quantum and classic. 335 00:42:52.550 --> 00:42:54.379 STFC-RAL-CR03 R61: and this is what they do for 336 00:42:54.780 --> 00:42:57.549 STFC-RAL-CR03 R61: so the variables that you see 337 00:42:58.010 --> 00:43:00.750 STFC-RAL-CR03 R61: on the x-axis is related to the 338 00:43:01.120 --> 00:43:03.290 STFC-RAL-CR03 R61: intensity of the process will be. 339 00:43:04.230 --> 00:43:12.700 STFC-RAL-CR03 R61: and according to to the results, a reasonable number for this density somewhere around here in those cases. 340 00:43:13.260 --> 00:43:19.530 STFC-RAL-CR03 R61: And then you see basically the threshold and all the methods compared. 341 00:43:20.210 --> 00:43:29.889 STFC-RAL-CR03 R61: So of course, when we are comparing quantum classical in high energy physics we I don't 342 00:43:30.040 --> 00:43:33.740 STFC-RAL-CR03 R61: expect to find that quantum will outperform classic. 343 00:43:34.310 --> 00:43:40.300 STFC-RAL-CR03 R61: If we had found that you would have heard about it by now, because we would be very proud of this. 344 00:43:40.660 --> 00:43:43.630 STFC-RAL-CR03 R61: But it's already a very promising result 345 00:43:43.790 --> 00:43:48.350 STFC-RAL-CR03 R61: that phantom can be progress on classical, and it holds up 346 00:43:49.080 --> 00:43:52.920 STFC-RAL-CR03 R61: so in this case, of course, they did find that it was very promising. 347 00:43:53.190 --> 00:43:59.170 STFC-RAL-CR03 R61: but it's also because a. The experiment is smaller. That's for tracking stations. 348 00:43:59.900 --> 00:44:06.350 STFC-RAL-CR03 R61: And we they applied A, so Q approach. They divided the problem in branches. 349 00:44:09.000 --> 00:44:12.119 STFC-RAL-CR03 R61: No wanted to take. 350 00:44:12.730 --> 00:44:16.019 STFC-RAL-CR03 R61: I always spend a lot of time talking about traffic construction. 351 00:44:16.340 --> 00:44:21.519 STFC-RAL-CR03 R61: but there are also many other areas in experiment and high level physics for being 352 00:44:21.700 --> 00:44:23.440 STFC-RAL-CR03 R61: kind of like on the moon with it. 353 00:44:23.980 --> 00:44:32.560 STFC-RAL-CR03 R61: one of which is being so. Of course, quantum computing machine learning, quantum machine learning. This is a really 354 00:44:33.430 --> 00:44:35.769 STFC-RAL-CR03 R61: interesting field. It tries to combine 355 00:44:35.880 --> 00:44:38.539 STFC-RAL-CR03 R61: the strong points from both sides. 356 00:44:40.130 --> 00:44:46.903 STFC-RAL-CR03 R61: depending on what you got. 1, 1, 2, you try to leverage more from one side or the other. 357 00:44:47.490 --> 00:44:52.620 STFC-RAL-CR03 R61: And what we tried at Latv is, we have this very common problem. 358 00:44:53.140 --> 00:44:59.710 STFC-RAL-CR03 R61: which is jets of particles, particles emitted in a corner like 359 00:44:59.850 --> 00:45:02.149 STFC-RAL-CR03 R61: the solution that looks like this. 360 00:45:02.930 --> 00:45:13.025 STFC-RAL-CR03 R61: And we had B and signets that we needed to identify. This is what we call it system like that. 361 00:45:13.790 --> 00:45:16.260 STFC-RAL-CR03 R61: So we wanted to know if yet 362 00:45:17.070 --> 00:45:22.469 STFC-RAL-CR03 R61: Contained a hundred phone dial by A, B and a C. 363 00:45:23.590 --> 00:45:30.250 STFC-RAL-CR03 R61: Now this email like the customer. 364 00:45:30.390 --> 00:45:34.780 STFC-RAL-CR03 R61: well, with neural networks and machine learning. That's machine learning. 365 00:45:35.330 --> 00:45:44.305 STFC-RAL-CR03 R61: What we tried to do here was to use quantum computer quantum computing and quantum machine learning, and then with variation and quantum classifiers. 366 00:45:45.406 --> 00:45:51.222 STFC-RAL-CR03 R61: This is a link of a different use case for technical section, because. 367 00:45:52.236 --> 00:45:56.610 STFC-RAL-CR03 R61: you don't need as many queries, because in this case the queries are. 368 00:45:56.960 --> 00:46:00.859 STFC-RAL-CR03 R61: take the information of the variables in order to train the algorithm. 369 00:46:01.180 --> 00:46:07.000 STFC-RAL-CR03 R61: So unless you have a really high amount of virus, you know, we cannot find 16 qubits 370 00:46:07.990 --> 00:46:24.200 STFC-RAL-CR03 R61: and also in quantum machine learning. Apart from computational complexity timing, we also sometimes try to see if the algorithm can pick up on correlations from the variables that we use to train your algorithmically. 371 00:46:24.490 --> 00:46:32.120 STFC-RAL-CR03 R61: And also if it could perform similarly with a less amount of variables or less amount of people. 372 00:46:32.740 --> 00:46:37.739 STFC-RAL-CR03 R61: because sometimes this will be the potential advantage from quantum efficiency lacks. 373 00:46:38.310 --> 00:46:41.050 STFC-RAL-CR03 R61: In this case we tried with 374 00:46:41.290 --> 00:46:43.719 STFC-RAL-CR03 R61: the new tab, which is the 375 00:46:44.080 --> 00:46:48.761 STFC-RAL-CR03 R61: the classical method. Also in the morning. 376 00:46:49.780 --> 00:46:55.000 STFC-RAL-CR03 R61: then you have the classical languages. So I'm giving you a network in this case, and then 377 00:46:55.420 --> 00:47:02.239 STFC-RAL-CR03 R61: a variational quantum classifier with 2 different encodings. So encoding information in 2 ways. 378 00:47:02.390 --> 00:47:05.099 STFC-RAL-CR03 R61: either in the ammo, the face of the pubic. 379 00:47:05.360 --> 00:47:07.479 STFC-RAL-CR03 R61: or we have to do for the queue. 380 00:47:08.610 --> 00:47:13.140 STFC-RAL-CR03 R61: Okay, so you can see that the glass monitors that split the 381 00:47:13.620 --> 00:47:18.990 STFC-RAL-CR03 R61: we call it 51%. 382 00:47:19.810 --> 00:47:25.180 STFC-RAL-CR03 R61: And the customer did a very impressively good, considering that it was a 3.rd 383 00:47:25.380 --> 00:47:26.969 STFC-RAL-CR03 R61: There's the thread. 384 00:47:28.030 --> 00:47:47.079 STFC-RAL-CR03 R61: This was the 1st result from quantum computing that we produced at the history, and there has been a lot of progresses that I cannot show you yet, because it hasn't been published. So it's only internal information. But basically it goes along the lines of this was b 1. Let's try b versus C, 385 00:47:47.670 --> 00:47:53.229 STFC-RAL-CR03 R61: and also something I kind of alluded to before is the entity status. 386 00:47:53.720 --> 00:47:59.329 STFC-RAL-CR03 R61: because when you have a quantum circuit you can compute the environment like entropy and receive it. 387 00:47:59.820 --> 00:48:04.354 STFC-RAL-CR03 R61: And if you have built your set within a specific way. 388 00:48:05.360 --> 00:48:13.770 STFC-RAL-CR03 R61: you can try to relate this entropy to the correlation of the particles on, or to the feature importance in this case. 389 00:48:14.404 --> 00:48:21.129 STFC-RAL-CR03 R61: No, you need to be very careful about this, because it really depends on the way you get your 390 00:48:21.240 --> 00:48:26.440 STFC-RAL-CR03 R61: the partitions in the periods. It depends on which students are kind of the experience. 391 00:48:27.170 --> 00:48:29.760 STFC-RAL-CR03 R61: and it depends on the interactions that required. 392 00:48:29.880 --> 00:48:39.490 STFC-RAL-CR03 R61: But there's a lot to be learned about trying to assess how the entropy of a quantum circuit could yield information on the actual values. 393 00:48:39.860 --> 00:48:50.179 STFC-RAL-CR03 R61: This is an example of trying to use the entropy for- for many other applications. There's a 394 00:48:50.280 --> 00:49:08.151 STFC-RAL-CR03 R61: there's another result that tries to use entropy in the quantum circuits for a normal introduction with one of the fundamental reasons why people want to look at quantum computing is because 395 00:49:09.387 --> 00:49:21.510 STFC-RAL-CR03 R61: correlations in a classical simulation, we are approximated, but in a contributor they could be naturally reduced. Of course, like this is already 396 00:49:22.360 --> 00:49:26.509 STFC-RAL-CR03 R61: Platonic idea in an ideal world. But it can actually lack. 397 00:49:27.550 --> 00:49:28.250 STFC-RAL-CR03 R61: Okay. 398 00:49:28.640 --> 00:49:37.470 STFC-RAL-CR03 R61: Now, what are we working towards now? Both personally, that's a common 399 00:49:38.391 --> 00:49:45.558 STFC-RAL-CR03 R61: from the experimental side. We seem to be leaning towards hybrid. 400 00:49:46.330 --> 00:50:05.383 STFC-RAL-CR03 R61: and I think this makes a lot of sense, because, of course, we have a lot of tools there and given the time scale. And the way we, the way with the data, it makes sense that we try to make the tools be also, considering the variation 401 00:50:05.920 --> 00:50:08.080 STFC-RAL-CR03 R61: desperate to try that. 402 00:50:08.760 --> 00:50:16.379 STFC-RAL-CR03 R61: So we have some interest. Of course you have to see how to best move information, because, as I said. 403 00:50:17.803 --> 00:50:38.190 STFC-RAL-CR03 R61: tell you, oh, you need to know the angle between the second. This needs to be passed to a quantum computer. Mistakes are normally your amount of time, especially if we are discussing 40 million times per second. So you know, this takes a lot of research. 404 00:50:38.520 --> 00:50:44.970 STFC-RAL-CR03 R61: because quantum quote can be more green and cold. 405 00:50:45.080 --> 00:50:49.746 STFC-RAL-CR03 R61: of course, is really heavily depends on the technology. I cannot promise you that 406 00:50:50.650 --> 00:50:58.742 STFC-RAL-CR03 R61: 1,000 qubit will. 1,000 qubits will not take up anything but the 2 main games, as far as I know. 407 00:50:59.440 --> 00:51:04.940 STFC-RAL-CR03 R61: is that either you need less units. And then, yeah, it's less. 408 00:51:05.260 --> 00:51:11.540 STFC-RAL-CR03 R61: It's okay, depending on the technology. Or you know, you think lasers. 409 00:51:12.348 --> 00:51:17.219 STFC-RAL-CR03 R61: So you know, there's there's some consideration there. 410 00:51:18.340 --> 00:51:29.000 STFC-RAL-CR03 R61: And then, of course, try to go for a small scale experiments. So 411 00:51:29.350 --> 00:51:36.220 STFC-RAL-CR03 R61: these experiments, especially the ones that are coming now with, with 412 00:51:36.380 --> 00:51:39.099 STFC-RAL-CR03 R61: whether somewhere in the green pipe or in the pit. 413 00:51:39.370 --> 00:51:43.130 STFC-RAL-CR03 R61: That would also be a nice where a place where to test this. 414 00:51:44.358 --> 00:51:51.381 STFC-RAL-CR03 R61: And then personal remark personally, think that for this algorithm 415 00:51:52.160 --> 00:52:00.046 STFC-RAL-CR03 R61: it's a very powerful algorithm. But we really need to think about different use cases because traffic construction doesn't seem to be to be the answer. 416 00:52:00.730 --> 00:52:05.858 STFC-RAL-CR03 R61: although I'd like to be told otherwise, because I devoted a significant amount of time 417 00:52:06.970 --> 00:52:12.555 STFC-RAL-CR03 R61: and then experiment simulation. It just seems like very good place for the top, right one to machine learning 418 00:52:13.070 --> 00:52:16.000 STFC-RAL-CR03 R61: like quantum grants and all sorts of things. 419 00:52:16.450 --> 00:52:21.590 STFC-RAL-CR03 R61: precisely because the correlations you can simulate, and all of this. And then once again. 420 00:52:21.870 --> 00:52:27.230 STFC-RAL-CR03 R61: when I'm saying all of this. I'm kind of not discussing what to do with the output 421 00:52:27.340 --> 00:52:29.389 STFC-RAL-CR03 R61: which we can absolutely discuss. 422 00:52:29.560 --> 00:52:35.019 STFC-RAL-CR03 R61: But I'm not doing this because it's quite a common problem in the rest of the function computing. 423 00:52:35.280 --> 00:52:42.790 STFC-RAL-CR03 R61: You can think that there may be phantom traps in the future to take your output. You can think there will be another building block 424 00:52:43.060 --> 00:52:44.429 STFC-RAL-CR03 R61: to take your output. 425 00:52:45.080 --> 00:52:52.450 STFC-RAL-CR03 R61: or you can try to find a smart way to take it out. Is that basically the 3 approaches people take. 426 00:52:53.460 --> 00:53:11.639 STFC-RAL-CR03 R61: Okay? So this brings me to my conclusion. I hope I have showed you that quantum computing is has a lot of potential for high energy physics. But we need to be very careful about the way to think about it, especially when it comes to data calling. 427 00:53:11.910 --> 00:53:19.870 STFC-RAL-CR03 R61: you know, because the moment we start to talk about implementing, then we need to interact with industry. And they. 428 00:53:20.430 --> 00:53:23.509 STFC-RAL-CR03 R61: I mean, it makes sense specifically. But I'm not thinking 429 00:53:24.230 --> 00:53:33.890 STFC-RAL-CR03 R61: we have a lot of products with building blocks, also from inside. So there's a lot to be back there. 430 00:53:34.110 --> 00:53:37.620 STFC-RAL-CR03 R61: And one product in my head we've been talking about because working with. 431 00:53:37.790 --> 00:53:45.709 STFC-RAL-CR03 R61: he's very rapidly expanding. Did you want to get in touch with me to be part of the Scriptures? And I hope we are 432 00:53:46.150 --> 00:53:48.839 STFC-RAL-CR03 R61: welcoming everybody that has any questions. 433 00:53:49.740 --> 00:53:50.620 STFC-RAL-CR03 R61: So thank you. 434 00:53:57.030 --> 00:53:58.510 STFC-RAL-CR03 R61: Thank you, Miriam. 435 00:53:58.650 --> 00:54:01.495 STFC-RAL-CR03 R61: If we can take some questions, you 436 00:54:03.150 --> 00:54:16.069 STFC-RAL-CR03 R61: you did all the possible segments, and you know the order in which you input the segments to the query Array. So you know, the 1st period is this segment. The second period is this segment. You know the order beforehand, because you have 437 00:54:16.650 --> 00:54:17.140 STFC-RAL-CR03 R61: they. 438 00:54:18.320 --> 00:54:34.039 STFC-RAL-CR03 R61: What you need to to pass is fixed to the, to the algorithm is a matrix that basically tells you the angle between the segments. So if the sentence, the coefficient one from the matrix, otherwise it's 0, basically. 439 00:54:34.400 --> 00:54:39.619 STFC-RAL-CR03 R61: So this matrix is very big. If you had 8 qubits, it would be 8 times 2 8. 440 00:54:40.340 --> 00:54:41.959 STFC-RAL-CR03 R61: But it's very sparse. 441 00:54:43.110 --> 00:54:51.080 STFC-RAL-CR03 R61: Then you have another matrix which basically is analyzing segments that belong to the same. 442 00:54:51.360 --> 00:54:55.290 STFC-RAL-CR03 R61: That chair and heads of sales, which I think you know because of the other need. 443 00:54:56.820 --> 00:55:01.300 STFC-RAL-CR03 R61: Yeah, so it's it's a matrix that will have alpha 444 00:55:01.450 --> 00:55:05.699 STFC-RAL-CR03 R61: if they share heads of things and 0 if they don't. 445 00:55:06.460 --> 00:55:14.460 STFC-RAL-CR03 R61: because this alpha comes with a positive sign, which means that when you minimize heavily, people like it, so you will try to make those disappear 446 00:55:15.710 --> 00:55:16.620 STFC-RAL-CR03 R61: that makes sense. 447 00:55:16.980 --> 00:55:20.829 STFC-RAL-CR03 R61: And then you have this end, which is, you can also completely beforehand 448 00:55:21.000 --> 00:55:27.100 STFC-RAL-CR03 R61: another parameters. So basically, if you have 449 00:55:27.580 --> 00:55:31.689 STFC-RAL-CR03 R61: 2 tracks, 3 segments. The number of active segments should be 450 00:55:32.750 --> 00:55:34.980 STFC-RAL-CR03 R61: things, or something like this to you. 451 00:55:35.120 --> 00:55:39.530 STFC-RAL-CR03 R61: These are half golden to 2 year old enough, you know, the only of the segments. 452 00:55:39.840 --> 00:55:47.080 STFC-RAL-CR03 R61: You know these metrics. You find a way to write this in terms of quoting gates. 453 00:55:47.480 --> 00:55:55.660 STFC-RAL-CR03 R61: and this is easy, because, since segment equals equals. B, it's actually a matter of politics. 454 00:55:56.060 --> 00:55:58.719 STFC-RAL-CR03 R61: All images in the same direction 455 00:55:59.100 --> 00:56:11.449 STFC-RAL-CR03 R61: is this, but that also means it's really a lot. Because if you have. Each segment has to talk to each other, and you have 8 segments. You can only imagine how many you can actually populate. It's a really good presentation. 456 00:56:12.762 --> 00:56:19.290 STFC-RAL-CR03 R61: I think that's it. Was that your question? Yes, yes, thanks. 457 00:56:21.540 --> 00:56:48.969 STFC-RAL-CR03 R61: Yeah. So one thing on the tracking is, 1st of all, it's very interesting work. You said they were straight, just to clarify. So there's straight lines. Yes. Is there any sort of obviously in a realistic system? They would be bent. So is there any sort of roadmap ahead for that? Or is that like A, is there any idea how to solve that just they have done some stuff with variation. 458 00:56:50.890 --> 00:56:54.470 STFC-RAL-CR03 R61: I think you have to try. But I think that 459 00:56:54.770 --> 00:56:57.978 STFC-RAL-CR03 R61: depending would actually be not so problematic, because. 460 00:56:58.830 --> 00:57:14.130 STFC-RAL-CR03 R61: you can try to relate, relate the penalty terms to the basics of the because, for example, you could relate them to the, to the scheduling, that, and at some point, of course you have to assume they are not different tracks, so I think. 461 00:57:16.475 --> 00:57:27.010 STFC-RAL-CR03 R61: Tell me, I thought with triplets, you know 462 00:57:27.170 --> 00:57:29.418 STFC-RAL-CR03 R61: you can allow for the lending to happen 463 00:57:29.900 --> 00:57:38.560 STFC-RAL-CR03 R61: more easily. The problem is that scalability really exploded. So that was reverse. Okay, thank you. 464 00:57:41.491 --> 00:57:44.400 STFC-RAL-CR03 R61: So early, you were comparing 465 00:57:44.560 --> 00:57:58.970 STFC-RAL-CR03 R61: an algorithm in on a quantum computer. I'm assuming that was on a classical computer simulated. I think it was previous. Maybe. 466 00:57:59.350 --> 00:58:08.260 STFC-RAL-CR03 R61: or at some point you compared the speed and said, Look, we're going to speed up 467 00:58:09.160 --> 00:58:37.369 STFC-RAL-CR03 R61: this one. No, no, it was earlier. At some point. You said, this took 2 days to simulate, yeah, yeah, the classical. Yeah, the quantum. Yeah, okay, so that was a quantum algorithm on a classical thing. Yes, not a classical computer trying to emulate them. Yeah. Good point, no, you have? So you have the classical computers. Yeah. 468 00:58:37.370 --> 00:58:45.607 STFC-RAL-CR03 R61: I refer to them as classical because they run distribution. What they are actually doing is emulator or simulated. I mean 469 00:58:46.030 --> 00:58:47.980 STFC-RAL-CR03 R61: a quantum computer. 470 00:58:48.354 --> 00:58:54.189 STFC-RAL-CR03 R61: They are run classical. You can run. It's a bio library. You can run them on your classes. 471 00:58:54.820 --> 00:59:00.459 STFC-RAL-CR03 R61: But yeah, you're actually trying the online. 472 00:59:00.590 --> 00:59:12.029 STFC-RAL-CR03 R61: So how much can you trust that when when quantum computers become feasible, you can actually run these algorithms on an actual quantum computer, how portable is the speed that 473 00:59:12.450 --> 00:59:13.510 STFC-RAL-CR03 R61: what was the 474 00:59:13.720 --> 00:59:33.919 STFC-RAL-CR03 R61: the efficiency of the algorithm is going to be the same when you port it to an actual good question. No, it's it's actually quite tricky when it comes to that. Because, take, for example, I don't have a picture here, but they have a very specific topology. Not every qubit is connected to every qubit. This is impossible. 475 00:59:34.050 --> 00:59:50.039 STFC-RAL-CR03 R61: So basically, you have your simulator, and then you have to what they call transpire it to the actual, which depends a bit of on the optimization that you choose. But they will try to accommodate, you know, to 476 00:59:50.470 --> 01:00:01.879 STFC-RAL-CR03 R61: to find the- the most optimal units what to put where and and try to accommodate your algorithm to the problem. But the problem is, if you have. 477 01:00:02.300 --> 01:00:08.520 STFC-RAL-CR03 R61: say you have Quebec one and qubit. 2 are not actually connected in the 478 01:00:08.670 --> 01:00:12.439 STFC-RAL-CR03 R61: but in your algorithm you need to. One stop to do 479 01:00:13.080 --> 01:00:25.310 STFC-RAL-CR03 R61: it can be done. But it will cost you more things, but of course, your- your 480 01:00:25.470 --> 01:00:31.340 STFC-RAL-CR03 R61: it's beautiful. You can do it, but it will cost you an increase in your system. 481 01:00:34.880 --> 01:00:41.280 STFC-RAL-CR03 R61: Any questions from the room, any questions from Zoom 482 01:00:41.700 --> 01:00:49.509 STFC-RAL-CR03 R61: in terms of the timeline? Is it? Or in terms of industrial deployment? Is it to try one of these algorithms 483 01:00:52.130 --> 01:00:58.349 STFC-RAL-CR03 R61: depending on people are really confident of themselves. I'm a bit more hesitant. 484 01:00:58.945 --> 01:01:04.619 STFC-RAL-CR03 R61: I say again, because I basically work with with them more but basically 485 01:01:06.760 --> 01:01:19.559 STFC-RAL-CR03 R61: they seem to be moving in towards they want to release like a ha 1,000 qubits, and they seem to be moving towards Kpus, or distributed systems connected to each other. This is already complicated. 486 01:01:19.690 --> 01:01:27.499 STFC-RAL-CR03 R61: and then the community as it is, and is moving towards a hybrid. You know where agencies and companies have to talk to each other. 487 01:01:28.070 --> 01:01:45.749 STFC-RAL-CR03 R61: It is it is feasible. But the way I see you can do 2 things. You can either try to access private funding computers and run on them, which costs a lot of money, and also not really because all this information has to travel. 488 01:01:46.030 --> 01:01:56.849 STFC-RAL-CR03 R61: And then there are some groups. There are some groups actually in Spain as well, where they have their Hpc system. And we try to build the quantum computer as physically close as possible to it. 489 01:01:57.300 --> 01:02:10.420 STFC-RAL-CR03 R61: which, of course, comes with a price, because noise is not as well isolated, and quantum computers superconducting quantum computers don't like heat and Hpcs, you know. So 490 01:02:10.570 --> 01:02:11.799 STFC-RAL-CR03 R61: it's tricky, but 491 01:02:12.300 --> 01:02:19.269 STFC-RAL-CR03 R61: it's visible. But you really need to think about how to transfer the information at that point. 492 01:02:23.040 --> 01:02:32.175 STFC-RAL-CR03 R61: And I know you wanted to give talk really, Agnos, agnostically. And you did. And now you don't talk. Can I ask? Do you have a preferred technology for cube? 493 01:02:32.920 --> 01:02:33.710 STFC-RAL-CR03 R61: Good. 494 01:02:34.320 --> 01:02:45.300 STFC-RAL-CR03 R61: The meeting is recorded now, but they cannot see my hand, can they? I am superconductive. 495 01:02:46.550 --> 01:03:00.668 STFC-RAL-CR03 R61: basically. No. But I, okay, is also doing a very, very impressive work in queue, and I think the school actually perform much better than we try to control the noise so definitely this to keep an eye on 496 01:03:01.753 --> 01:03:08.250 STFC-RAL-CR03 R61: superconducting me. It's easier to understand for me. Obviously, without coming from a hardware background. 497 01:03:09.300 --> 01:03:11.949 STFC-RAL-CR03 R61: we- we have to see how it goes. And also 498 01:03:12.220 --> 01:03:26.240 STFC-RAL-CR03 R61: some algorithms are not native to some quantum computers, you know. So actually, one thing didn't say to all of this, I would say No. 499 01:03:26.400 --> 01:03:32.320 STFC-RAL-CR03 R61: actually, if I tried that much better, because it's much more worthy. 500 01:03:32.890 --> 01:03:33.580 STFC-RAL-CR03 R61: Thank you. 501 01:03:35.530 --> 01:03:36.330 STFC-RAL-CR03 R61: Sure. 502 01:03:36.820 --> 01:03:43.319 STFC-RAL-CR03 R61: if no one, then I maybe I can ask question for this carrying algorithms, which we did. 503 01:03:43.964 --> 01:03:51.510 STFC-RAL-CR03 R61: Did you tried? So there are 2 questions to that. One thing, how much percentage of the algorithm was running on the classical computer. 504 01:03:52.144 --> 01:04:01.579 STFC-RAL-CR03 R61: I'm I'm I'm certain that the you know, the preparation part and the later on interpretation part might be in the classical. How much percentage of all that, and the second is 505 01:04:01.820 --> 01:04:07.350 STFC-RAL-CR03 R61: daily time measuring throughput or evaluating throughput and scalability. In this case. 506 01:04:07.830 --> 01:04:19.510 STFC-RAL-CR03 R61: Yeah, interesting. So the percentage well, we have physics, data from. And so I think that's very. 507 01:04:19.920 --> 01:04:24.572 STFC-RAL-CR03 R61: That's all. Your values have already been completed by you, the difficult assets. 508 01:04:25.070 --> 01:04:34.199 STFC-RAL-CR03 R61: because it's a variation quantum, same classically, to find the optimal values. 509 01:04:34.420 --> 01:04:51.190 STFC-RAL-CR03 R61: And then for the output, the think, we actually an actual iphone computer back then, not possible. But 510 01:04:51.510 --> 01:04:52.510 STFC-RAL-CR03 R61: and we'll save it. 511 01:04:53.380 --> 01:04:54.330 STFC-RAL-CR03 R61: 18. 512 01:04:54.590 --> 01:05:19.949 STFC-RAL-CR03 R61: The input is completely construction. You don't need anything collapse, you can find a way to store your information handful of qubits and just collapse those to measure those. So that's easier. And the scalability is also easier, because 16 qubits was already a realistic amount of features to train well enough. 513 01:05:20.420 --> 01:05:24.105 STFC-RAL-CR03 R61: So scalability today will keep constant. 514 01:05:25.210 --> 01:05:28.760 STFC-RAL-CR03 R61: and when you study entropy, you will have that final level of. 515 01:05:29.160 --> 01:05:36.150 STFC-RAL-CR03 R61: But you need like oxygen of this, and let me see great 516 01:05:36.823 --> 01:05:39.970 STFC-RAL-CR03 R61: any other question from the room or zoom. 517 01:05:42.860 --> 01:05:46.600 STFC-RAL-CR03 R61: I see no hands raised up, so let's thank our speaker once again. 518 01:05:51.610 --> 01:05:55.450 STFC-RAL-CR03 R61: So thank you. If anybody wants, we'll get.