Einstein Telescope Technology Project - Decode the universe's secret with Artificial Intelligence

Einstein Telescope Technology Project - Decode the universe's secret with Artificial Intelligence

Show Video

Okay, good afternoon and welcome all of you here at the brightland Smart Services campus and all of you watching from home or at the office. And today we are going to talk about a very exciting Topic at least as far as I'm concerned the Einstein telescope. My name is Martina hermans and I am honored to guide you through the project today or to this to this event. Honored I actually mean that because we are going to talk about two of my favorite topics. So artificial intelligence in combination with the Einstein telescope.

And a little bit about the the Einstein telescope. I mean we had over Einstein who predicted that there was something called gravitational waves and Professor Hilt will tell us all about what that is in a minute and he predicted that in 1916 and it took us about 100 years to prove that he was right and well Professor hildew who's with his today was also part of the research team that discovered that and Now we want to know more. So now we want to build an even bigger and even more accurate research facility. And we want to do it here in this region on the border of the Netherlands of Belgium and of Germany. And that is what we call the Einstein telescope. We want

to build it deep under the ground and we will hear what that will look like and how it will work. And I say we want to because it's not certain that we will get it we have competition and the competition is Italy. And so in 2024, we will know if it will be located here of or in Italy and then it will take quite some time to actually build it but I think the lookout is very exciting to have such well potentially such a very high impact research facility here. And it's not just research as we will also here today from Renee. Kesse from Leo and also from Roy from boosting Alpha. It has a much wider impact in society and in business than just in the research community. So we're

very excited here today to tell you all about that. There are preparations going on. It's a massive project a massive undertaking and one of the preparation projects is called the Einstein telescope Technologies project. And that is what we are working on today. And we would like to show you what is happening there. This is the first session of in total four sessions in which we will catch you up on what is going on because what is going on is it's pretty cool.

So we would like to share that with all of you and we would also like to invite you to think along and to ask questions or to come up with suggestions things. We might not have considered yet. So all of you here, but all also all of you online. If you have any questions, you can type them in the in the chat function. I will get them here and we will answer and discuss as much as we can during this during this session.

So to start I would like to invite Professor Stefan Hilt to the podium. He is Professor of experimental physics at the University of masticate. He's also the head of a research facility called the ET Pathfinder in which we experiment with Hardware that could be used to build the Einstein telescope and he will explain to us. What is the Einstein telescope? What are we trying to measure and how are we trying to do that in the Einstein Technologies project? Yeah, thank you very much for the nice introduction. Welcome everyone.

And yeah, as you heard I have the pleasure today to talk a little bit about the Einstein telescope, but mainly about this project to give an introduction of what is ET Technologies in particular. What are the work packages in addition to what we will hear and much more detail later today. I just need to see how that works. okay, so we heard already gravitational waves Einstein predicted them and we are really really fascinated by the new window that we now have open to yeah, listen to the customers listening to the mergers of black holes and nutrients are maybe in future even hearing Yeah, starts blowing up Supernova and things like that and the key here is really to understand what is in the cosmos what drives the cosmos what are the forces in the cosmos and we know most of the energy in the universe is actually not consisting of meta and like like us like what we have in our hands and so on but these terms Dark Energy Dark Matter 95% of the universe is made up of these and we have very little idea of what this is and obviously, you know a scientist that's quite frustrating if you know that there's so much out you don't know so this is really kind of you know, the big picture what we want to do with gravitational wave.

So you're really just at the starting point here many people compare to where Galileo was when you know, he built up the first telescope and look at the moons of Jupiter. Good. So that's the science pitch in all of this and we had heard before we want to build an Einstein telescope. So what's so special about the Einstein telescope right? Now. We have some marvelous facilities three to four kilometer big gravitational wave detectors, which were actually constructed in the 1990s. We were working on these for you know, I work my entire career on these machines here and these are marvelous.

They brought us the first detection of gravitational waves, but the problem with them is they're really not sensitive enough for what we want to do. So to give you an example, we want to start our next science one actually in February next year. Then we will switch these into from it as on there and then for every week that we have these operational we think we will hear about one second of a signal And that is yeah, it's nice. It's many many more signals

and the 100 years beforehand, but it's still not what you really want. So with the Einstein telescope, we want to have a game changer in here and instead of occasionally hearing some signals over the noise more or less. You will hear signals all the time millions of signals and actually is a challenge turns around rather than actually waiting for the few signals to arrive you have too many cigarettes, you need to disentangle them and so on and that's obviously part. I think we will hear a bit more about later today good. So that's the Einstein telescope marvelous facility and obviously lots of Technology involved and that's one part why we're here and why we are talking with industry. There's pushing technology on all different fronts if you want to build it's actually detector and if you want to analyze the data again talking about, you know collaborations with boosting Alpha with the universities who do data analysis.

I think we will hear a little bit more about this but this is really why not only scientists get interested about this, but why we have Year of you know, the government engaging in many many companies and that's super useful because also we couldn't build it on our own scientists can build small stuff in the lab but scientists are not going to build a two billion Euro facility underground and stuff like this. So a very useful Synergy here. Why are we so interested in the Einstein telescope? I mentioned a few of these points. It's really kind of, you know, all of these new technologies give you kind of you know, this Innovation radiating out into industry attracting people educating the next generation of scientists, but then you know, we see it from from the people we have in the universities most of them go to asml and to other companies and actually help the economy and this is really kind of what drives us here to have the Einstein Center the scope actually quite nice close by here. Good in order to get to that there's now quite a big ecosystem that has been built up here in the region with different projects a few were mentioned already. We started

off. I think the first one was ET pass finder where we have a small version of the Einstein telescope to test technology in a very similar environment to the Einstein telescope and the incredible sensitivity and there's etest EMR. This is testing two different sets of of things one is about the geology mapping the geology. Where do you actually build this? How does it work with underground and so on and the other one is testing a particular cooling technique for silicon mirrors in there.

We have ETS to SME and you see there is a collaboration with industry is really in the foreground where small companies can actually work across the border get vouchers for certain Technologies to develop these and actually one of our partners here. I think jpe for example was also an ETA test ET Actually, one of the recipients of such vulture to develop a new device. Then we have ET Technologies and we'll talk about this more. There's a national growth funds 42 million again for development of the ecosystem working with industry, but also the underground and then there are more European ones where we set up the project office for the Einstein telescope. So obviously we are always working on two different fronts here. We heard that we want to host Sia

instead and Telescope here. So that's obviously one one Circle in which we work and so on but as a scientist, I also work on the other fronts are much bigger front is obviously we want to have one Einstein telescope in Europe and that's obviously also something where we actually help developing this and then there are things like the Dutch black hole Consortium. So quite a big ecosystem which covers all the different aspects some of them more focus on the region some of them more focus on the European thing, but all hopefully trying to push forward that we soon will have an Einstein telescope. So now we come to ET Technologies. What are the main purpose of this particular part of the ecosystem? And here we want to do two things which we have here two main objectives one is developing an improving Technologies. Yeah, there are certain

things if we talk about vacuous system infrastructure underground and so on where really and many of the of the cryogenic Technologies, we're really the biggest skills are actually not only in research institutions, but also at industry and obviously want a profit from from the synergy. So this is really kind of making the answer telescope possible technology is better and at the same time while working together, we can actually boost the regional economy prepare actually companies to understand, you know, what is actually needed to then when the big Investments come being in a very good position to bid for this money. So these are more lesser two objectives. And then this is focused in six. It's also

an organization of work package but six real topic work packages. I will quickly go through and I have to say it's quite fascinating we're saying to someone earlier today. I don't think I've ever been part of a project which is so diverse in terms of the range of topics, but also in the approaches it people do in different work packages. Some are very much focused on a very narrow task others are much wider to look at what is actually possible and really kind of scoping out and it's a little bit, you know, it's not clear as a journey will go and so on so actually quite interesting to see this So what are we doing? So on the cryogenic side and always put the companies and the institutions here in the upper right corner, and this one which applies mainly to ET pass finder where we want to learn with cryo world with jpe how to actually cool a mirror in a very stable way for this gravitational wave detectors. You need to have very quiet mirrors. If you connect them to cooling system cooling system always doesn't noise. So the question is how can you in between

find either way that the noise from the cooling system is not transmitted or that in between you have things you actively control and stabilize to keep them quiet. And this is what we do actually in this work package one. For work package 2 and we are thinking about the vacuum system. So the Einstein telescope as it's designed right now. We require 120 kilometer

of vacuum tube will be the largest vacuum system in the world and they're funny funny comparisons. If you just scale how these three and four kilometer detectors were building you do it at the same speed and so on then just the construction of the vacuum system for et would take 25 years. Okay, and that's obviously something that you don't want to have here you need to do this more efficient you want to save money and so on and there are quite a lot of interesting techniques how to make it maybe cheaper by using different types of Steel applying certain Coatings to protect the system but also to improve its performance and vacuum and here has teamed up together with a steel producer Tata and vdl obviously do quite a lot of these processing controlling and detailed planning. Work package 3 is inner size which is a spin-off company from from nikav, which does census to measure what vibrations are there in the ground. So they have very nice seismometers that we actually use in campaigns all over the place and also in the current detectors, but they now want to look at gravity gradient sensor. So to look at how the field changes

over space and that's something which is quite hard actually to measure and there's some clever ideas how to actually do this and that is currently explored in this work package. Work package 4 again. You see we're jumping all over the place. It's very very different topics work package 4 is nikaf and Oracle Netherlands. So this is kind of using satellite data.

And some some clever actually, I think patent and protected algorithms to use surface data. So you scan the surface with a satellite the height and so on and from that actually predict the underground structure rather than having to go and drill holes and takes a course out and so on they have some magical waste to give you information what happens there water flows and unstability and we are currently looking into this and I hope over the next few months. We will see the first results what you can actually learn with these magic Technologies. Work package 5 I will keep very short because I think that's something where we want to hear more later. Today is really

kind of just to understand there's lots and lots of interesting algorithms out there to do in a classical way. They turn analysis or classical trading and I'm pretty sure there are many many more clever algorithms out there whether it's networks or whatever and your parts you have and that would be quite interesting to understand how you can improve or if you can improve with them gravitational wave. Data analysis by data analysis just to be clear always means two parts in the past. We were mainly after finding a signal in

there. But now the computational power is mainly in this part of once I have my signal I want to learn about the physics from the object. Where does it come from and that is obviously a slightly different challenge in here, but I think they are useful applications for both of these. And then the final work package actually led by Rob who is here by today. It's about sustainability in the wireless sense of sustainability. Not only the ecological thing but also about kind of you know, how can we make effects of the positive effects of the Einstein telescope will have a lasting imprint on the region here. So it's really

kind of you know about the Economic impact as well as kind of the quite a lot as obviously environmental impact if we think about this nice region here, what do you do with all of the stuff you take out of the tunnels? Where do you put it? How do you get it away without building a Motorway here to get it actually out of the way. What happens about Water Management? If you actually build something underground, you probably need to pump, you know Waterway, which otherwise would get into your tunnels. What does that to the water levels here the water that you collect you probably have to put somewhere up again. Can you just put it into the room? There's quite a lot of interesting aspects here that we need to think about and the whole topic of energy and so on good and I think with that I think I yeah, thank you.

But in particular, I think also everybody who sponsors this project and who made this project possible and in particular all of the partners and as I said, I hope that was okay to give you a really kind of a top level overview. It's a very diverse project and about each of these work packages. I think we can talk for half an hour if people want and I hope at coffee will have some occasion for that. But today obviously it's all about the work package five and I'm curious to hear what people have thoughts on these Technologies. Thank you very much.

Thank you very much Professor. Stephan Hilt. Are there any questions? Sorry, I was wondering which of the aspects of all the technological developments will actually make the resolution or the sensitivity so much better than the other existing telescopes. I think there's two different aspects here and one is actually that it's a moment. Our gravitational wave detectors are built to cover a whole frequency range. So like if you think about light the whole color space and

so on why we now already know from astronomy. It's much easier if you think about astronomy right now, you have completely different telescopes for x-rays for optical light for infrared light and so on and the same we are trying to do with the ants and Telescope. So one thing is to say instead of building one interferometer to cover all of the frequency range you split it into two but it's so called xylophone and that gives you already quite a lot of sensitivity because you have different noise sources which dominate different parts that you can actually in your spectrum and that helps a lot and the other biggest thing I think is that we go cryogenic that one of the limiting noise sources.

Is that everything wobbles what has temperature? In particular your mirrors the elements from which you hang your mirrors and so on and by cooling that to minus 250 Centigrade that I think will give you another big big Improvement and then there's another 10,000 small things, but I think these are two of the of the biggest ones I would say. Thank you any more questions. No. Thank you professor.

Then we move on to to our second presentation of today, which is by Renee cassif from Leo. He works as a business developer and he will talk a little bit more about linking this seemingly, very scientific and ever to Industry and to businesses particularly small businesses and how we can benefit actually economically also as a region in the businesses in the region from this Einstein telescope. You Martina? Thank you. Okay as Professor Stephen hilt already mentioned there's a lot of things going on already in the context of the Einstein telescope and the decision for location still has to be made in a few years. But as we speak there are a lot of developments already going on a lot of Technology developments going on and a lot of opportunities are included in these developments. But these opportunities business opportunities, they don't reveal themselves by themselves. They don't develop themselves by themselves.

And this is why in 2021 seven Partners in the region, they cooperated and set up a project to build a bridge between the technological and the business opportunities on one side and the companies in the region on the other side. And this is what's called the project ET to smes which started in March 2021. My name is Renee gesson, I work as a business developer for leeoff and Leo of is one of the project Partners in project ed2smes and it's also the project partner in the Dutch region of the interch EMR region where this project will take place. Now what I'm going to talk to you is about let's say the summary of this project what we do in this project. And also what are the goals of these projects and

how we approach that? And our targets are is to include and involve all the companies in the region EMR in the developments that are already happening at Pathfinder at etas and a lot of projects. And help them to find the business opportunities specifically for their business. And what we also want to stimulate is cross-border. Collaboration between companies between research institutes and universities smes and large companies cross border in this region and to let's say to grab the opportunities in in these developments.

And what we also would like to emphasize is that the developments in the ions and Telescope they are not just to use within the framework of the item telescope all the products involved in Einstein telescope all the Technologies development. They also can be applied in other markets. And there is only one big ice in telescope later. So the volumes for the Einstein telescope Market is not that much but if you can apply this technology or product in other markets like Automotive or Aerospace, then these are significantly large markets and also the business case is of course then much more interesting. Now to build to create a this bridge between Both Worlds, we have set up a complete value chain that consists of four big steps. We have the mapping as one step. We have business development support as the Second Step attract the fundings as the third step and sustaining as the fourth step. And in this presentation, I

will guide you through all these four steps. Now the first step the mapping part is what we did there like in the project ett. We also have let's say Defined 12 technology domains that are relevant in this Einstein telescope framework and this 12 consists of eight for the instrument itself.

And for for everything that has to do with the underground and the infrastructure. And these technology this technology catalog is on our website ET to smes and if you for example click on cryogenics, then you get a whole worksheet a whole paper on this specific technology and it describes what are the challenges in the eye institoscope context. What is the current state of the art? What are potential other markets for this technology? And this is done for all of the technology domains that are in this catalog. Now what's also in the mapping tool? We try to create a map where all the relevant players in this Einstein telescope where they are active and these are companies. These are research institutes. These are universities large companies smes and for every partner involved we have made a profile that describes who is the company. What is the core business? And

what is the link with the Einstein telescope? And so far we have created a map with already more than 300 relevant actors in This framework of of Einstein telescope. And of course, they're not all the same involved in this framework. The other company has more. Let's say more connection to the ions in telescope than the other one. This is the exercise that

we do to explore with each company. That is interested. Okay, where is the opportunity for you in this project? The second part of our approach is the business development support. Now what we have done in the beginning of a project we have send out a survey over all of our networks in the region informing companies about the Einstein telescope what will happen and what is already happening and we ask them basically the simple question. Are you interested in developing or discovering the opportunities for you in this project? And based on the response. We have conducted company visits with each company that expressed this interest we sat together. We got to

know the companies we explained what is happening in the Einstein telescope and looking at the current business and their products and Technologies. We try to discover. Okay, what would be the connection with your company in this project and where and which way you can benefit for everything that is happening in this project. And so far within the region we have visited 217 companies the visits are still ongoing and will be ongoing during the course of the project. And what we all did is we connected the companies to

the specific technologies that we that we defined. And so far 344 competences where related to the Einstein telescope catalog. Now what we also did you have to tell the story? Of course? A lot of people when we started this project and actually nowadays. It's still the case many people have not ever heard of the Einstein telescope. If you talk to people businesses today, some people say eisentaloscope. What's

that? So you have to inform them about what is the ions and telescope? And what we also do is to let's say to be visible in for example fares or exhibitions where we participate as a project team either by visiting the event and spreading the word there or by participating with a booth. For example, we attended the Precision bureaus in in desert organ boss last year. We will do the same this year. And we also visited relevant technology related events technologies that are somehow related to the Einstein telescope. And we also invite companies that say to to discover the possibilities the opportunities of the Einstein telescope there. What we also did in this work package is we organized some online and physical workshops. We had

planned to conduct six physical workshops originally end of last year, but for apparent reasons and they could not take place. So we decided to have two online events on instrument Technologies and on geology Technologies, and we moved the physical workshops to the period April until June of this year. So we had six workshops also a bit divided over the different Technologies. And

actually the last Workshop was last Tuesday and there we not focused on the Einstein telescoping Technologies, but we focused more on what it takes to let's say to keep the Einstein telescope running. For the next 50 years once the start button is is Switched. And we focus there on maintenance on health and safety and also energy Supply and Water Management. Now the third step in our project is that we have attractive fundings.

And if as a company you see the opportunity to develop a product or process or a technology. That is somehow related to the Einstein telescope. And you also have a partner in the region. Across border that wants to collaborate with you had to fill a certain competence Gap or to let's say to add some some expertise then you can apply for an innovation voucher that I say takes away the first risks the first costs of this Innovation project. And we have grants up to 50,000 Euros to support this Innovation project with lead time of up to nine months.

Now what's important is that this collaboration has a minimum of two companies. Cross-border of which one at least one is an SME and the other one can be an SME or can also be a larger company. And additionally beside those two companies you can also have research institutes universities or whatsoever involved in this project as a partner as a supplier. Now we already had five calls where applications can be submitted and the next call the sixth call in this program opens on August 1st and closes on August 29th of this year.

Now if you look at these Innovation projects, I already mentioned that we challenge companies not just look at the Einstein telescope as a potential Market, but also look at current markets other markets where you can use this product or this technology and what helps is that the Technologies in Einstein telescope, they perfectly fit within the key enabling technologies that we have defined here. And so if you look at the instrument Technologies, they perfectly match within all the areas of the key enabling Technologies, and also the geology Technologies, they can be divided somewhere over these areas. So Auto potential markets could be astronomy identical but also Aerospace and other more automotive and also other relevant markets that you can see as potential interesting for this project.

now how we approach this with companies that have a certain project idea. They contact us they give us a call. They send us a mail and they say okay have this project idea and I would like to see what's possible and at that time it's just an idea. So what we

do there is we help the companies to Define this project ID in a five-step. Let's say format. So that starts with okay, who are you and what are you doing right now? What is the project idea? What is the Innovative news? What is the link to the Einstein telescope? And what kind of partner do you need in this project? What kind of competence are you looking for? Now after this, they filled in this format, we will discuss this project idea in our business development team and that consists of business developers from the seven participating partners. And then we're going to look in the other regions what potential companies what potential Partners can be found can be linked to this competence to this expertise that is needed by in this project idea.

And if if we have a match then we arrange introduction meetings where the companies can get to know each other. They can discuss the project idea and then during this this meeting there will be yes or no a first match. Okay, this this there's some chemistry going on and there is indeed a technology match and a competence match and that is basically the start of let's say further development of the project and of the collaboration And if the project is defined and if the application is ready, then they can submit the application for a call in the voucher scheme. Now the business developers in this project. We have three on the German side. We have one in balloon

three and Flanders and one in the Netherlands. So we invite the companies in this region. Please contact your business developer. If you have a project

idea if you have been inspired by the opportunities in this Einstein telescope framework and you want to really grab these opportunities. The last step is what we call sustaining. Of course, it's an indirect project and it has an end date and we certainly do not want that all the activities all the achievements that we have achieved in. This project will end at this end date. So we're taking care that all the things

created. They continue after our project has ended. And one of the examples is that we have set up an industrial Advisory Board together with the project etpart finder and the project e-test. where scientists and companies meet physically or perhaps online to discuss opportunities to discuss technology developments and to to let's say to to find each other and not just a scientists and the companies but also the companies located across border and a lot of interesting matches already made between companies from one side of the border and companies on the other side of the border that decided to collaborate also outside the Einstein telescope context.

So it's really a good platform to expand your International Network and especially in the region. Now if you want to have more information. My name was here on the right side as a business developer, but you can also see some information on the website ET to smes.eu. We also have some promotional material Communications also one part of the project package and Yeah, this is the let's say these are the seventh partners that are included in in the project. We also have a few Associated partners that help us finding companies approaching companies located in a different areas. We also have partners that really support our project and they wrote a letter of support letter of intent by the application of the project ET to smes.

And of course a final word for our sponsors in this project indirect and also some parties within the region. And that is in a nutshell. what I want to talk about the project ET to smes and of course if you have any questions or remarks Now is the chance? Thank you. Thank you very much Guinea. I

see already one question. So one another institution participates to this project will help does it work in terms of funding? So the funding goes to the smes and the productivity the research institution can participate or also, the research institution gets funding out of this schema. Yes. Good question the project that we do we are subjected to the European rules on funding and these will say in this case that we can only fund the smes that want to innovate. We have collaborations between smes and larger companies and in this particular case, the developments are also interesting for the larger company not in terms of funding, but for let's say for a longer a long term strategy and they can use the things that are developed in this project for other purposes as well rather than just this project and the same also applies for research institutes that are involved in this project.

Thank you. any more questions Thank you very much, honey. You're welcome. We received the couple of questions through the chat and I would like to I think these are more for stay fun still. So maybe we can take a few minutes to to answer them shortly and then we will take a short break. So one of the questions is many of the partners currently are Dutch organizations. The question is will there be any International collaborations for example with MIT or with CERN? Yeah, very good question. So there are collaborations. So

we as I said that comes back to this kind of thinking about the difference fear. So the Einstein telescope is a scientific project which covers all countries in Europe about 1,200 scientists in Europe also has kind of a sister project in the US called Cosmic Explorer and therefore example, we interact quite closely with people Caltech and MIT and so on so I think that and actually their Partners in the Einstein telescope which are for Japan and from Australia, so that does not anything that is closed or so. Yeah, so that everybody is welcome. And CERN is a slightly different thing. So we have some Mo use and we are now starting up a few different collaborations on a certain topics.

So for example, the vacuum system is something we're now at the ET collaboration has made an mou with CERN and CERN will develop the long beam pipes together for example with you know Partners here from from work package. Two I think and so on so there will be also interactions with some because certainly has a very long history and quality control making big science project actually work. So it's a win-win situation. And yeah, we're collaborating with them. Thank you very much last question and will you or do you work together with the Event Horizon telescope team or any other teams or events that work in in black holes and and the results that come out of that? Yes, I think we are collaborating. I mean,

for example me personally we collaborate with a group at rutbau University High northalke who's one of the brains behind the Einstein telescope and we're writing grants together and so on so these collaborations happen and whoever has I mean the last few years have shown us was this dawn of multi messenger astronomy where we saw the nutrients Dam merger where then you know, we learned quite a lot of this event after the gamma ray telescopes and the gravitational assets. So to say triggers astronomy Community, they're more or less in all different parts of the astronomical community and different Spectra people started to observe this event and we saw how much you know how much information you can get out of this huge synergies. And yes, whoever is interested in collaborating on black holes. I think we are very happy as it. Thank you very much. Then we will take a 10 minute break. We will start again at three o'clock for everyone

who's watching online. Please feel free to also get a cup of coffee and keep the questions coming. We will return after 10 minutes. test test All right, welcome back everyone. So we will continue with our program and we will now move to work package 5 as earlier explained by a professor Hilt which is the work package about artificial intelligence and we will look at it from from two angles. So this work package is a collaboration.

Between the open University the University of Utrecht and nikaff and boosting Alpha which is a financial trading company located in venlou. So it's a also collaboration between industry and and the Academia and we will first hear from Liana koye who is an assistant professor at the open University from a scientific perspective what we're trying to discover with the help of AI in combination with gravitational waves research. Yeah the floor is yours.

Thank you. Just to have an idea before coming here who had an idea of what is AI? You have heard about it, but can you define it and who knows about the gravitational wave? Okay, just to know a little bit so as much as I said, I'm Liana. I'm a subdivision scientist working at open University. And today it's one of the first presentation in about the work package 5 where we are trying to deploy tools for detecting gravitational wave and get ready for when the Einstein telescope data will be there. So as Martin I said the work package is a collaboration between three sort of expertise expertise from the gravitational wave. So the physicist from the open University

and nikkev expertise on animal detection from boosting Alpha, but I will let Roy talk about that in the second part of the talk. And expertise on senile processing and deep learning from open University. and what we are trying to do is that we are trying to develop tours to find an unknown signal in existing data set but we are also trying to using the current graphical detectors and deep learning techniques create a synthetic data set which will look like are we expect the Einstein tells of data to look like so just to give you an idea and under the control of professor hit. Gravitational waves actually are Reapers in the otherwise tough and stiff space time Fabric and what we are what happened is that those reapers are caused by really violent phenomena which happened in the cosmos so you can think in terms of colliding black holes exploding stars or even the birth of the universe itself. The problem is that when we think of the detector of nowadays which were built in the 90s some of you were not even born yet.

we can think that the technology was not far advanced yet. So the data the completed of those asophysical events, barely. Pass the background noise of the detector. So the gravitational wave that we have observing right now are swamped in the noise in the background noise. And what happened is that in terms of signal processing what we do is that we do much filtering. So we are trying to

find the signal. into the noise, so I try to put it try to trivialize it but let's say that we have noisy signal like that We create a bank bank of templates and then we are going to compare our templates to our observation. And we evaluate what we call cross correlation. Where we try to look for significant matches, if we think that we have a significant match we say that we have detected an event. And the cause of the event is actually going to be determined by which template. Was used to identify. So we see but to

give you an idea if we think of my voice. We are that's also a wave. So we are thinking of a wave which is going to in a room like that.

If you have a womb like that, which is quiet trying to identify my voice. Is okay. It's a little bit of work, but you can do it. if I'm standing in a room all our Auditorium of Ola PhD defense become a little bit more difficult to pick up my single voice into and to figure out that I am the one once you have detected the signal to figure out that our yam the one talking it's even more difficult imagine that I'm standing in the middle of a BTS concert.

That's what is happening. That's what we are trying to do. So it's work which is tedious. Time consuming and computationally heavy. So what we are trying to do is

that we are saying what can we use? What can we do? to detect and to have a more Syrian and Parent Time pipeline to detect Gravitational wave you see me coming what we want and I will paraphrase and entering is that we want a machine which can learn. From experience. So we want something which is going to learn by itself to pick up the characteristic and the feature.

Of the signal. So this is called machine learning. This is a part of artificial intelligence and we are actually even using a subset of machine learning which is called Deep learning where we look into neural networks, which actually are trying to mimic and to learn but in really the human sense that you think of so it's really it's trying to mimic the process which is happening in the neocortex of human. So when you think of learning as a child or something, this is what we are trying to do apparently Google's succeeded there. Is this sentient AI over there. I don't

know if you anyway All that to say to you that we have reached. a world now where the new Industrial Revolution is actually driven by data. It is a data-driven revolution where you look into artificial neural network. and when we think of serendipity what we see is that by the time that the machine learning deep learning has been starting since the 50s.

But by the time that we started to have computational poor to be able to do that was actually the same time where we were being able to detect with the technology the first gravitational wave. a few heard of deep fake Yeah. actually I want to talk a little bit about dick fake because believe it or not deep fake is using techniques that we are going to use.

In this work package and I say we I'm not doing anything. This is the PHD student still funnel Tom Melissa. Those are the one really doing the work. And actually what they are doing is that they are using something that we call Guns generally generative adversial Network where the net it's a technique where you use two Network and one is called a discriminator. The other one is called a generator. and the generator what it does is that from a bank of data that we call the training set that you give him.

Is going to learn about the statistical feature of the data and is going to be able to generate new data. The first deep fake was for example Tom Cruise appearing somewhere yet never been to. That's what is happening. So you learn from data photograph. A data from interferometer spectrogram kilograms you learn statistical parameters and then you generate new data. And what you do is that you feed those data to the second Network, which is called a discriminator.

And you try to fool him to think that those are real data. And if that works, then you have create a good Network, which is able to generate new data. So if we are able to do that.

that means that we if we take the observation that we have currently and we are able to Generate that that means that we can do different things. We can do that that equality Improvement because we can generate new noise. We can generate what we call glitch or beliefs which are transient noise which comes so we know or today do them. So it's easier for us to just create a

bank. of those glitches and to provide them to the scientists. You can also use it for creating the gravitational waveform. The

gravitational waveform are basically the template I mentioned before where using the Theory let's see simplified like that you can Create exactly. Our astrophysical event would look like problem is is that it's computationally heavy. It's not something that you can simulation that you can run so easily but if you train a network to be able to do that, then it might take sometimes to train it. But one it's once it's trained you

can easily deploy it. The other way, it's also simply for graffitational Signal search. So for example, there are team right now, which are looking into using neuronal network to identify CBC. When two black body collides coalescence, so they form one so they are trying to look for that in the data.

And something that Professor is mentioned before also. Where you don't you're not solely interested in identifying those of observing way if you want to know what is before. What is the source and this is also where deep learning can help you because they can help you to interpret. What was the origin what was the source of the gravitational waves that you are observing? So now what is happening is that and this work package? We have a few students. So Stefano which

is sitting there Melissa, which I think join online which are looking into harnessing the power of the guns. So harnessing what you know, this deep fix to characterize and generate new population of glitches. and blips and we have Tom for example, which is looking at the same technique but to looking into new architecture to be able to generate artificial Collision population. And that's just a teaser because in November today, it was really to get you acquainted with what we do and for more General audience, but in November, we will have a more in-depth presentation where they will have the opportunity to really present to you in depth each of those applications. And I really hope that I will see you and that you will also spread the world that this is an opportunity. That's about all I had to say to you today.

Thank you very much Liana. Are there any questions for lyanna? one question Give me a moment and come to you. Yes, sorry register for this event to in November. Oh, yeah the same way as now you will receive we will advertise him advertise it. Sorry, and you will get a link and you can register.

Okay. Thank you. One question. You said you you are able to create new data sets for for et.

Kind of assume in the context of mock data challenges and so on so that means you you take so what exactly does it mean you take gaussian noise and then the part is you put these blips and glitches at you you created in there or do you also create the do something to the gaussian noise that you use? No at the moment. It's just started. So they are really just developing and training the algorithm to be able to train it. But ideally the idea it would be to be able to do a little bit. Like what we do in the satellite world to do observe observation observing simulation system experiment where you take contact with the hardware people and then they discuss about the project specification.

And then you can create real specific data. So I think at some point Tom will contact you or the other member of the VP so we know exactly What are the current achievement and that we can use as a PSD for creating really realistic data set, but we really just started in January. So If I can add something on that, we're also working within DT collaboration. There is a Mark that a challenge going on now. and oh, yeah and This is actually something something that. Okay, so yeah, we're part of that and we'll probably use the package that we are developing for for injecting glitches in there. Yeah.

any more questions No, thank you very much Liana you. So we got a little bit more technical there for those of you who wanted even more technical. We would we invite you to join the session in November.

But for this session, we would like to conclude with what in academic terms. We call valarization and valarization is basically Everything that we achieve in the world of science also has an application outside of the university and in this specific work package. We have one company called boosting Alpha who is not only learning from us, but also we are learning from them because Roy Landers is here and Roy is the chief technology officer of boosting. Alpha Roy is a successful entrepreneur has had many AI driven startups in the past and at boosting Alpha he together with his team is working on trading algorithms for financial markets. And those are actually very Advanced and it turns out that even in the world of physics we can learn from each other. So this Financial Training algorithms

have an inspiring effect on the scientists and also the science that's going on has an inspiring effect on the boosting on boosting Alpha, which is Rider in this work package. So we didn't want to keep that from you also today to show you real life application. They are using it. So Roy the stage is yours. Okay. Thanks Martina. So yeah, I'm going to have my name is hollandos Chief

technology officer of boosting Alpha. So what we do is we specialize financial markets and we build basically complex trading algorithms that try to find patterns to make money. All right, so that's what's happening. Why do

we do that? Why is it so difficult because everybody knows, you know, everybody has probably experience with Equity or Forex. You're trying to predict the stock price, but one thing is for sure and financial markets, whatever happens has happened in history will never repeat again into the future. So that makes it quite difficult to Predicts financial markets and it's also makes it quite difficult to apply artificial intelligence because you know that the data in which you train is never going to be the data of the future, right? So Let me see how this works. So what we specialize in basically artificial intelligence quantitative Finance algorithmic trading we use both financial data, but also non-ferrential data, so think about algorithms that trade based on Twitter feeds we we can use all kinds of different data sources. So what are we trying to solve? Basically in the world of financial trading there's many hedge funds there's many investment managers and every year there are reports being done. They have all the ad phones in the world basically and everybody is trying to beat the market right? But if you look at in reality how many hatch funds can actually beat the market it's very low. It's roughly 5% of investment managers

and headphones can actually have actually the capability to beat the market. The rest is basically if the NASCAR goes up 15% and your fund goes up with 12% Yeah, you could have better invested hundred percent of your money in the sun on the poorly and not in a hedge fund because they made a lot of Trades but apparently they the benefit from those trades were yeah, you lost money basically So that's why we're trying to do. So a little bit about a company we started in June 2018. And basically we focus on building the algorithms only so we don't bring them to Market. So we work together with B2B partners that actually use our arguments and sell them to Market. One of the partners is for example, the Bots platform. You

probably see them on on TV with ads. They're very active with marketing. They're actually trying to be a unicorn end of this year. It's a startup from Alam and roughly 60% of all your all the arguments are not platform. We

are the biggest trading algorithm provider. So so that's what we do. But we also work with growing mirrors in Netherlands, which is a big creator of chains with beersboro with miracle and Israel signal in the US.

So we were active in many different different parties. Um, yeah. So what are we trying to do? So we collect historic data hysteric data from financial markets. and let's compare ourselves with the manual Trader. If you are a manual Trader and you look at the stock charts these Financial charts and then people try to predict. Okay, is it going up going down? Right and

I meet a lot of manual Traders and sometimes manual traders that I meet that have device this intelligence strategy, right and with taking this signals and I made 100% in six months that can happen right those people exist. I actually met them they made thousand percent in six months, but I think that they can repeat that in the next three months. And once somebody is successful they always they want to earn more so they always raise the stakes but sometimes down the road after nine months. They go bankrupt. That's what usually happens. And why is that happened? Because a manual Trader cannot test his strategy on History, right? He just starts. And he does things on feeling.

But he's not not 100% sure that actually works and what we do with algorithms. We actually test our algorithms on 20 years 30 years of History. So we properly test our algorithms on Good Times on bad times on stable times on all the different types of Market environments that you can actually happen. We try to

test our algorithms to see how stable are they? You know, how did they do in different markets environments? And and we all also test life so usually when we tested on history and we have a good idea. Okay stable enough good enough this will you know and endure the test of Life data then we put on also three months. on the shadow Trading so three months on live data operating to see OK. Is it really, you know, did we not overtrain it? Does it really work? And then we basically start with real money. So what are we doing? Is this project? We are crazy guys, right? So we test a lot of crazy stuff and we are very experienced with AI. So one of the things we are bringing I think hopefully in this project is that we help the teams from Utah University and open University with some IDs on how you could apply AI to gravitational wave detection.

But what we are primarily interested in ourselves as well is can we learn can we beggyback can we learn some techniques that are used in gravitational wave detection that we can actually apply on financial markets and when we started this project we thought that would be two big interesting teams for us. One of them is using data from multiple exchanges. Why do we think that's an interesting idea because the the current gravitational wave detection as a state fund shown already. They use data from three different detectors. Why do they do that? Because those three detectors are spread across the world. So suppose the signal is only

picked up by one of the three then it's probably noise, right? Because if it's a real signal it would have been picked up by all three. That increases the confidence in the signal basically. And that's the same for us, right because many Financial assets are not traded on one exchange. They are traded on different changes could be one in the US One in China. Could be different volume different Market participants, whatever. so one of the things we are always looking is can we find this quick discrepancies right discrepancies, which are not logical so and that means we can benefit from it. Right? So the

other technique that that is interesting for us. The gravitational wave detection team is looking for in model what they call in model searches. They're looking for signals. They've never seen before so they don't know how it looks like, right? If I'm looking for the big bang for example in the Einstein telescope, nobody saw the Big Bang yet, right. So

how does it look like? And we are trying to do the same thing on financial markets. We are looking for patterns, but we don't know which patterns right? So that's an interesting area for us as well. So what are we concretely doing? So around this multiple exchanges thing. So we have Gregor's here in the room. So what we have been collecting and there's basically

live since March this year is from three different changes binance crack and coinbase. Those are three of the biggest crypto exchanges. Binance is primarily Asia. Carcan is us and coinbase is also us and we collect every single trade and every single change in the Autobook. So

whenever somebody in the world places an order, we collect it. So it has been running since March. So we're talking about terabytes of data that has been collected so far huge data volumes. And what we have been exploring is if we compare these this data on exactly the same time frames do we see differences? Right? Because if we see differences for example for the price of the Bitcoin If the price on binance is 1% difference from the price of coinbase, that's a big topic right 1% Is Big you with one trade we could earn one percent. And you would say okay the price is not on percent different. That's Bitcoin is a heavily traded access with a lot of volume. But yes,

we've seen occasions already where prices are 1% different for exactly the same financial asset. Now I'm talking about crypto. But shell for example shell is traded on the IX and enrollment. Exactly the same underlying asset.

But the prices could be different. Right. So basically because the underlying asset is exactly the same you could make money without any risk.

And that's obviously what we try to get right we try to find discrepancies make money without risk. So that's what we're the other thing we are trying to do is if we combine the signals from all three. Could we buy combining the signals? Could we could we create a higher resolution of the data? By combining the signals and reducing local noise, right because we might have trades that only only happened on I don't know binance. Which did not happen, there were did not happen to the auto exchanges so you could say it's local noise. Which did not get replicated to the other exchanges. So what we also trying to do is

if we combine data from multiple changes create a new data set with less noise and then run all traditional algorithms. Because then you get better quality data basically as input for the trading algorithms. That's obviously a little bit more the difficult technique. So we are trying a lot of stuff like Helms genetic algorithms outer encoders.

And trying to figure out we're working on this product together with University of mistake. So we also working with some professors trying to figure out techniques how we can combine data from multiple different places. To basically get to a higher quality data format. And the other one and model search is patterns.

So we learned from the gravitational wave detection team that the current telescopes work primarily with. The existing templates and all of those templates and I think Liana showed that are based on Einstein's relativity, right? So we know from Einstein's relativity theory that the signals should look like this. In our world, we don't know how to signal looks like but we don't we do know that our patterns. Now one of the things that Financial on financial markets we look for patterns is a technique from Japan called Japanese candlesticks.

So that's why we decided to focus on Japanese candlesticks because a lot of manual Traders actually look at Japanese candlesticks to so the typical traders in a manual traders in a trading room. They look at charts like this. To actual

2022-07-03 08:48

Show Video

Other news