AI trained on simulated galaxies helps astronomers correct noise and reveal dark matter patterns matching standard cosmological models.
Japanese astronomers have developed a new artificial intelligence (AI) technique to remove noise in astronomical data due to random variations in galaxy shapes. After extensive training and testing on large mock data created by supercomputer simulations, they then applied this new tool to actual data from Japan’s Subaru Telescope. They discovered that the mass distribution derived from using this method is consistent with the currently accepted models of the Universe. This is a powerful new tool for analyzing big data from current and planned astronomy surveys.
Wide area survey data can be used to study the large-scale structure of the Universe through measurements of gravitational lensing patterns. In gravitational lensing, the gravity of a foreground object, like a cluster of galaxies, can distort the image of a background object, such as a more distant galaxy. Some examples of gravitational lensing are obvious, such as the “Eye of Horus.” The large-scale structure, consisting mostly of mysterious “dark” matter, can distort the shapes of distant galaxies as well, but the expected lensing effect is subtle. Averaging over many galaxies in an area is required to create a map of foreground dark matter distributions.
But this technique of looking at many galaxy images runs into a problem; some galaxies are just innately a little funny looking. It is difficult to distinguish between a galaxy image distorted by gravitational lensing and a galaxy that is actually distorted. This is referred to as shape noise and is one of the limiting factors in research studying the large-scale structure of the Universe.
Training AI on Simulated Galaxies
To compensate for shape noise, a team of Japanese astronomers first used ATERUI II, the world’s most powerful supercomputer dedicated to astronomy, to generate 25,000 mock galaxy catalogs based on real data from the Subaru Telescope. They then added realist noise to these perfectly known artificial data sets, and trained an AI to statistically recover the lensing dark matter from the mock data.
After training, the AI was able to recover previously unobservable fine details, helping to improve our understanding of the cosmic dark matter. Then using this AI on real data covering 21 square degrees of the sky, the team found a distribution of foreground mass consistent with the standard cosmological model.
“This research shows the benefits of combining different types of research: observations, simulations, and AI data analysis.” comments Masato Shirasaki, the leader of the team, “In this era of big data, we need to step across traditional boundaries between specialties and use all available tools to understand the data. If we can do this, it will open new fields in astronomy and other sciences.”
Reference: “Noise reduction for weak lensing mass mapping: an application of generative adversarial networks to Subaru Hyper Suprime-Cam first-year data” by Masato Shirasaki, Kana Moriwaki, Taira Oogi, Naoki Yoshida, Shiro Ikeda and Takahiro Nishimichi, 9 April 2021, Monthly Notices of the Royal Astronomical Society.
DOI: 10.1093/mnras/stab982
Never miss a breakthrough: Join the SciTechDaily newsletter.
92 Comments
Science is to know the unknown.
The title is misleading. The work is about correcting the distortions from gravitational lensing in the observable universe, thereby determining the correct shape of imaged galaxies. I cannot find anything about resolving the shape of the universe.
It resolves the local curvature involved in the lensing.
There is absolutely nothing about this that has anything to do with the “shape of the universe”. This is just a filter for clearer pictures.
If you read the paper, you can see from the equations that there is a connection to local (weak lensing) shape.
So, what is the shape of the universe?????
The universe looks on average 3D flat space, but you get that from the large cosmological data surveys when modeling the cosmology.
Yeah. And a CERTAIN Worcestershire Sauce has 57 herbs and spices. But it can’t change a lump of bologña into a filet mignon. It surely WILL make a 4oz. Kyoto Beef filet taste like BALONEY. That’s a personal opinion… MAYBE a personal preference. It’s ALL in the palate; TASTE, in other words. This study TASTES “funny”; as in FUNNY-STRANGE. TO WIT:
A priori: “reasoning” which proceeds from theoretical deduction rather than observation or experience. AKA: my guess is as good as yours.
A posteriori: Reasoning from facts to general principles. SOMETHING like: I THINK therefore I am… what I THINK I am.
Both approaches are flawed. The common “thread” is perceptual BIAS. Wedding the two (as does this STUDY) creates a DOUBLE FAULT which ultimately SQUARES the degree of DEVIATION; particularly when ONE arrives at “conclusions” which END where PRESUMPTION began.
But OF COURSE (the PATH to CERTAINTY), adding a SUPER COMPUTER (OMG! It MUST be true) is the current test of CREDIBILITY. It is more: a DISABILITY, inasmuch as: recapitulating a priori assumptions (@ hyper-speed, no-less) in a MODEL will INEVITABLY lead one to conclusions “consistent with THE (emphasis applied) current cosmological model.”
It “flys” in Principle, but not in Practice.
By this “model” I PRESUME you mean (((THE BIG BANG!!!))). Get over it. “IT” is BIBLICAL in scope and its prognostications, AND Ego-centric to boot: POINT SOURCE (god), interacts with THE VOID; gives rise to EVERYTHING (which is MORTAL) AND, like the HUMAN BODY, “IT” corrupts over TIME; shrinks in on itself and ultimately RETURNS to the “dust” from whence it came. It is symmetrical, in a sense, but I won’t bet the FARM on it; although I will certainly BUY THE FARM one day… at a TIME.
As a pilot, one accepts this POTENTIAL or ONE keeps both feet PLANTED on the ground, to sprout (maybe SPOUT) weeds of intuition in the stead of EXPERIENCE: It is not about FLIGHT as much as it is about FLYING. HOWEVER, forget the PHYSICS of the former and the latter will terminate at GROUND ZERO: You buys the Farm, but you will never own it. In ACTUALITY, there is risk inherent in this activity: Difficult to see that errant downdraft spilling off the downwind side of that cumulonimbus formation…
One gets the picture… The BIG PICTURE in a controlled spin. Practice makes PERFECT!?
WHAT? You might ask, does THIS have to do with THAT? Answer: CONCEIVING THE HEAVENS. BIG BRAIN MEGA DATA PROCESSORS “know” nothing about the Universe, Life or what it is to BE. You can PRETEND as much, but: SHOW ME.
Not going to BET the farm or BUY the farm on IT’s advice. I, for one, need to FEEL the DIRT before I EAT IT. Computers will never feel the way I feel: GOOD.
Define GOOD.
Sorry, HAL. Just messing with your “MIND”. It IS 2021 (or thereabouts), not 2001. Here, at the “cutting EDGE” of SCIENCE, nothing can go WRONG… go wrong… go…
WRONG! I am, therefore I think. You AIN’T. I’ll just pull this here PLUG…
🦋
APOLOGIA: I tell myself that I will make my point in 100 words or less. But thoughts are much like jelly beans: difficult to stop with just one. Orders of Magnitude, here on this Independence Day: A FIFTH of July to go…
@Alex You’re supposed to point fireworks away from your face.
What @Alex suffers from is called “Diarrhea of the mouth; constipation of the ideas.”
You sound like Dan Winter.
BIO-ACTIVE Systems Dan Winter was one of the first scientists to provide evidence that the shape of the universe is a stellated dodecahedron
Actual shape my rear. Stupid title. This is a THEORY, not a fact.
A well tested science theory is a fact.
You are thinking of some other word use.
Bit disappointed as the headline made me think they’d figured out the topology
They make the title of the articles sound like a huge discovery so that people will get interested and click on it, but they’re actually lies.
They title may be lies at times. This one was not (they model local shape).
Looking at the depicted view of the universe…looks like a tray with a folded piece of paper towel that the one side is curled up. Reminds me of when germinating seeds in spring. And all the microorganisms in the water that live on the seeds and and paper and in the tray…what an interesting concept if you compare us and we are the micro organisms on those seeds in that tray with the paper towel…representing our universe the water being matter all around…lol
I’m no scientist but where is the picture they have of the shape of the universe. I don’t see the shape they have observed. Did I miss something? I want to know what is the shape of our universe…. 🌍🧿
a donut
Yes – you didn’t read the paper, which explains what they did and how it connects with shape of (local space).
So is it flat or curved? Oh wait the article has nothing to do with that.
This is was a waste of time to read through. It explains little to nothing. It pretty much states that they simulated a bunch of useless tests. Trained the A.I. to do something that has not even been done by humans yet.. I get all excited about the solar system, galaxy, then the big one the UNIVERSE.. question though how can something that can not be fully seen due to positioning, obstructions etc. Ever be truly mapped out unless there is a way to get around these obstacles… Don’t tell me cosmic radiation and such because there is far too much of that to trace the web work out.. However this read was as useful as my comment.
The article describe method progress. That will make us get quicker to the goal of understanding.
Why are people hung up on this issue of “shape”? If matter distorts/shapes space, then a technique that sheds light on the dark matter distribution in an area of the universe would necessarily shed light on the shape in that area of the universe (and vice versa). What else is a shape other than a collection of gauges/measures on something’s parameters/dimensions?
The shape is flat. I find it extremely funny the article is written to say this newest AI method just reasserted the already held belief based on the current accepted model of the universe, that its flat. Not open nor closed. Good job AI..
It’s like you’re complaining that the latest LIGO experiments are confirming General Relativity.
The only good thing that came from this was reading the comments… Thanks
I think Alex is Dan Winter.
So wtf is the shape of the universe idiots
COMPLETELY MISLEADING!!! Dishonest title This is why am stopping reading your articles is good information but why you have to be dishonest with what it’s.
It wasn’t misleading – what mislead you is that you didn’t read the paper for understanding.
My hats off to Alex who explodes like a supernova from a singlarity so simple and seemingly unexpandable and unimportant. But as the onion peels away you begin to see his explaination. Humans want to be more. Which road leads to more? Our universe being a shape this saying there could be more/alternate shapes or flat out…were in space and that’s that… Humans it seems will go for the path that CAN be travelled.As we grow ever more closer to the are we a simulation layered in a.i. with multiple other layers of simulation around us? Or are we real with limitations yet to be reached? Or are we beings limited only to the next level of questions and answers in an endless cycle of discovery? The shape of the universe I theorize is a vehicle In which to picth the minds of human kind forward from the mundane flat earthers ideas to the idea that we are definitely more and have barely scratched the period at the end of this sentence about who we are.<=== (For closed minded people please end all life at last period see "<===") but for those enchanted by the chance that we are more and the word learning is just that and it never ends please by all means keep moving forward strive to be more till you either tire of it because you are in fact more or you tire of the rigamarole of questions and answers. Hint real believers in the quest to seek who we are and why will never stop. So in that I applaud this as a significant achievement for those who buckle down and strive to be more than some human on some crummy planet disturbing the flow of time with meaningless existence….I always say….if you're going to do anything …extend all the energy you can into it…if energy is neither created nor destroyed by doing what you love with all you have you will live forever and achieve immortality. In summation….believe that you are more or you will be nothing.
Oh my…… My first question is are you AI? My second response is yes, yes, and yes yes… Your answer is correct, true to some and not to others…… You make it seem like it’s something wrong with striving to be more than what something or someone thinks that you are not, which is wrong for you to say that because everything strives to be better to understand to know more.. as long as it for goodness and to get to know your higher power even more, to get closer to our higher power,but, you know that’s not what they’re going to do that’s what makes it sad. They move to take the credit out of our higher Powers work, that’s what makes it really sad, but us that has eyes to see and ears to hear this work proves of our higher Powers existence even more. My thing is will they feed us a bunch of lies while hiding the real truth only feeding us bits and pieces of the real truth just as they didn’t show us the picture of the universe they claim they have? You’re right, humans can be something else. 😆😂🤣 May the most high continuously bless me and those wh….
Please capitalize all MOST HIGH HOLY AND ALL HIGHER POWER in this conversation I DEEPLY APOLOGIZE I did not out of REVERENCE MEANING RESPECT TO OUR CREATOR OF ALL THINGS.
I’m disappointed that, after reading this article, I still have no idea what the shape of the universe is. But, I have enjoyed reading the comments section. So, not a total loss. 🙂
I didn’t know the eye of Horus was the Galaxy, most interesting. Yes we want to see the picture of what they have of the universe.
21 square degrees out of a potential 41,000 is a relatively small sample size, but still A massive amount of data. Intriguing, but we have a long way to go.
Alex does not need to “Drive”he is crazy. There is NO one in outer space ppl .
https://hubblesite.org/contents/articles/gravitational-lensing#:~:text=Gravitational%20lensing%20not%20only%20distorts,size%20of%20the%20galaxy%20cluster.
With gravitational lensing there should be a way to find there is no end.
Is there a shot where we could lense to another galaxy doing the same
It seems they meant the CONDITION of the universe, and not its form, such as if seen from a distance. It is shaped like a torus as I reveal in a book I am writing. Remember the name Kevin L
They meant the local shape – but the larger universe is 3D flat space on average (seen in large surveys).
Shut up
Why!? Trolls bore me. If my comment bores you, don’t read it.
And be polite if you can. No reason to attack other people if you have trouble with reading or understanding.
Actually my preceding comment about the universe being shaped like a torus (due to the fact that it IS A TORUS) was posted at about 12:40 AM EST on Tuesday July 6th 2021. I did not know this site would post my surname, so yes, Kevin Landry is my name.
As I said earlier, the universe is shaped like a torus because it IS a torus and then someone with this site deleted or hid it and wrote “a donut” as is the shape of a torus as if it was their own ideal. Caveat: Thieves and deceivers do not inherit the Kingdom of God and are by Him punitively recompensed as I invoke Him to justly do to whoever further deletes or hides this comment. 2nd Thessalonians 1:6… Repent!
yeah, a donut
that was July 5, 2021 at 1:32 am
that was July 5, 2021 at 1:32 am
Good to see you repented. Now get yourself a hard copy Bible and use e-sword.c o m free Bible study software download and start feeding yourself spiritual food and eventually you will discover that the universal torus concept is the means whereby faith and science have always been, as it were, “kissing”… And then some!!!
Irrelevant superstition.
“…there is no ONE in space ppl…”
_____
While I am tempted to just say either “Sources or disregard” or simply “prove it” I will refrain and opt for the more civil:
“Not only does that run astronomically against the odds, but it is also information that you could not possibly possess assuming you are not some deity or an equivalent higher dimensional being. You may want to back off the stating speculation as if it were fact in the future.”
So I know little more than before I opened this article.
Kevin Landry is spamming.
Dunces :!!!
Take as many DODECAHEDROns as you like Multi Verses and add a Dollop of Annamorphic Squeeze Factors and by and Large if not more or Less You will end up With a Pizza Fit For A Druid at Beltane…Don’t ask me What’s on the Pizza as it’s only 3 seconds into the Big Bang!
Dunces :!!!
Take as many DODECAHEDROns as you like Multi Verses and add a Dollop of Annamorphic Squeeze Factors and by and Large if not more or Less You will end up With a Pizza Fit For A Druid at Beltane…Don’t ask me What’s on the Pizza as it’s only 3 seconds into the Big Bang!
Preprint here: https://arxiv.org…2890.pdf .
“An ensemble learning technique with our GANs is successfully applied to reproduce the PDFs of the lensing convergence. About 60% of the peaks in the denoised maps with height greater than 5 sigma have counterparts of massive clusters within a separation of 6 arcmin. We show that PDFs in the denoised maps are not compromised by details of multiplicative biases and photometric redshift distributions, nor by shape measurement errors, and that the PDFs show stronger cosmological dependence compared to the noisy counterpart. We apply our denoising method to a part of the first-year HSC data to show that the observed mass distribution is statistically consistent with the prediction from the standard ΛCDM model.”
This methodology complements the basic AI use in gravitational lensing, 1 of 4 such astronomical uses – planet hunting, gravitational waves, finding dynamical events and gravitational lenses – that has been reviewed recently [ https://thenextwe…dication ].
Apparently it suffices to model weak lensing as a first derivative shear matrix (eq 1), which components in turn can be related to the second derivative of the gravitational potemtial.
The primary use is of course observational, but it is also informing on cosmology:
“Although we assume the WMAP9 cosmology in the training process, the denoised PDF by our GAN shows a cosmological dependence as shown in Figure 8. Hence, it is not trivial if the denoised PDF for the real HSC data is consistent with the WMAP9 cosmology. … This indicates that our GANs extract some cosmological information hidden by observational noises.”
Small gains, but since adding dark matter effects from neutrinos, i.e. using a νΛCDM model, in a recent paper gave a 1-2 % improvement, such percent level improvements become interesting in our new high precision (< 1 % uncertainty) cosmology:
"Our GANs can reproduce the one-point probability distributions (PDFs) of noiseless fields with a level of 0.5−1 sigma level. This argument holds even when we vary the multiplicative bias with a level of 1%, the photo-z distribution of source galaxies, and the error in galaxy shape measurements with a level of 10%."
Neutrino physics is hot, I read that the European ESS neutron source that will be used for material study may also be used to generate neutrinos and send them north through the rock to deep mines in Sweden and Finland. Neutrino physics promise to explain the matter-antimatter asymmetry, and it would give us a leg up on similar efforts elsewhere. More ν ("nu") physics! 😉
Preprint here: https://arxiv.org…2890.pdf .
“An ensemble learning technique with our GANs is successfully applied to reproduce the PDFs of the lensing convergence. About 60% of the peaks in the denoised maps with height greater than 5 sigma have counterparts of massive clusters within a separation of 6 arcmin. We show that PDFs in the denoised maps are not compromised by details of multiplicative biases and photometric redshift distributions, nor by shape measurement errors, and that the PDFs show stronger cosmological dependence compared to the noisy counterpart. We apply our denoising method to a part of the first-year HSC data to show that the observed mass distribution is statistically consistent with the prediction from the standard ΛCDM model.”
This methodology complements the basic AI use in gravitational lensing, 1 of 4 such astronomical uses – planet hunting, gravitational waves, finding dynamical events and gravitational lenses – that has been reviewed recently.
Apparently it suffices to model weak lensing as a first derivative shear matrix (eq 1), which components in turn can be related to the second derivative of the gravitational potemtial.
The primary use is of course observational, but it is also informing on cosmology:
“Although we assume the WMAP9 cosmology in the training process, the denoised PDF by our GAN shows a cosmological dependence as shown in Figure 8. Hence, it is not trivial if the denoised PDF for the real HSC data is consistent with the WMAP9 cosmology. … This indicates that our GANs extract some cosmological information hidden by observational noises.”
Small gains, but since adding dark matter effects from neutrinos, i.e. using a νΛCDM model, in a recent paper gave a 1-2 % improvement, such percent level improvements become interesting in our new high precision (< 1 % uncertainty) cosmology:
"Our GANs can reproduce the one-point probability distributions (PDFs) of noiseless fields with a level of 0.5−1 sigma level. This argument holds even when we vary the multiplicative bias with a level of 1%, the photo-z distribution of source galaxies, and the error in galaxy shape measurements with a level of 10%."
Neutrino physics is hot, I read that the European ESS neutron source that will be used for material study may also be used to generate neutrinos and send them north through the rock to deep mines in Sweden and Finland. Neutrino physics promise to explain the matter-antimatter asymmetry, and it would give us a leg up on similar efforts elsewhere. More ν (“nu”) physics! 😉
Bogus comment. Read bait.
Methinks that the comments section be worth the read and not the article. 😄
The title didn’t fool me. The size might come in handy before determining the shape.
Iff the universe is ever expanding what are the law s of boundaries??
General relativity, which underlies LCDM cosmology, has no need for boundaries – the modelk describes all of space without it whether or not the universe is flat or some ither shape. But there are way to describe various boundaries such as the so called cosmiz horizon of the observable universe for an observer [ https://en.wikipedia.org/wiki/Cosmological_horizon ] – different boundaries for different observers.
There is also specific boundaries for topologies that may be popular for various reasons (such as ease of calculations). AFAIU if a so called de Sitter space of theoretical physicists has a boundary, it’s vacuum energy will have a net contribution from that. There are so called “holographic” boundaries for string theories. Et cetera.
A well tested science theory is a fact? That is not true. A theory, even tested, is still just a theory. Theorys tested are tested in controlled environments and conditions, can’t really but that to the real world.
That is a fact. 😉
You are describing the common usage of “theory” outside of science, but what we now have to call “science theory” to distinguish it from such opposite notions is the oldest,
“theory (n.)
1590s, “conception, mental scheme,” from Late Latin theoria (Jerome), from Greek theōria “contemplation, speculation; a looking at, viewing; a sight, show, spectacle, things looked at,” from theōrein “to consider, speculate, look at,” from theōros “spectator,” from thea “a view” (see theater) + horan “to see,” which is possibly from PIE root *wer- (3) “to perceive.”
Earlier in this sense was theorical (n.), late 15c. Sense of “principles or methods of a science or art” (rather than its practice) is first recorded 1610s (as in music theory, which is the science of musical composition, apart from practice or performance). Sense of “an intelligible explanation based on observation and reasoning” is from 1630s.”
[ https://www.etymonline.com/word/theory ]
A modern science theory can be described by statistical test methods used to establish observational facts. Only you can derive multiple predictions and do more tests, so they are broader and more robust – like how “an image say more than a thousand words” – I like to call them “superfacts”.
I’m not sure how to interpret your last sentence, but that – seemingly saying that lab doesn’t test on the real world – as well as the bait-and-switch with common “theory” usage looks like taken out of creationist playbooks. You may have a problem looking for sources on science [I suggest Wikipedia, at 20 % error rate it is far better than creationist 100 %].
So what is the shape of the universe then?
Yeah … as of 2018 Planck collaboration cosmology paper they identified the universe as 3D flat (to the best of all the then data – which flatness has only become better evidenced in later surveys). It was a repeat, but a crucial one since they had managed to filter out earlier noise.
Bias in any system derived by human mind will inevitably lead to the conclusion inherent to that system
Says your own bias.
What says the evidence? Very little bias in cosmology, for example.
Space is Fake. A complete man made construct. No one has ever been or will ever go to outer space. Low Earth Orbit is the furthest humans has ever gone. It’s all we can do.
And people literally believe there are remote control cars and helicopters on Mars. That notion is absolutely bonkers. Cartoons, CGI, and images (imagination) are what NASA, and other world space organizations, are made of…
If space is fake, you are fake.
I think the time stamp of sundog speaks volumes. 452am.
Truly amazing. Truly.
@Torbjörn Larsson if it does honor its title then by all means please SHOW us the shape of the universe……
See the illustration beneath the title.
It shows it twice over, since the suggested average flatness can be derived from the cosmic background image that is topmost.
Fascinating..but calling ML an AI, and local observable space the Universe is wildly inaccurate…and I’m surprised they just started using machine learning to do this..shouldve been done long ago IMHO…
Crap.
As a Born Again Christian with Wisdom and Understanding, I can tell you that the Universe is a Globe similar to the Earth. Understanding is a flat plane, and Wisdom is a Spirit Woman.
So, irrelevant superstition is your thing?
Not the thing in science.
Knowledge is what is learned.
Wisdom is how effectively that Knowledge is applied.
Wow. Lots of opinions. I wonder how many are actually scientists in with peer reviewed publications. Just curious…..Sci-curious
Ahem. But not in cosmology.
Also irrelevant, not only because it is anecdotal and you would want statistics. At any given time, the rate of scientists in society interactions is ~ 0.1 % globally and 2 % in Europe. So say that every other of these many commenter pages is visited by a scientist (at the very least, since – well – science!).
It isn’t odd, if that is what you are asking. You meet a scientist, assuming you and your partner isn’t one, every walked city block.
This title is absolutely misleading, no matter what the finnish panther is trying to convince you of in the comments … They input and dissolute material generated through relative observation. This does not define the shape or dimensions of our universe, rather theorizes the shapes and dimensions of the constituents within our universe based on relative observation. Lens shaping is but one artifact deviation being pondered when battling extreme distance observations and their conclusions are but the premise for hypothesis.
All will be revealed.
What is the difference between the shape and the actual shape? What meaning does actual add to the article title?
Torbjorn Larsson, you are the biggest loser I have ever seen. I only commented because of your gross need to comments on everyone’s comment who finds this odd. There is no AI or human that knows the shape of the universe, honestly don’t be such an idiot. We don’t know.how big the universe really is, we don’t know how many galaxies lie beyond our own solar system, so how can we possibly know the shape of the universe. Stop talking down to people and trying to sound smart. Your just a basement dweller in grandmas house.
ARGENTINA 1 – BRASIL 0. THAT’S MORE IMPORTANT
I would assume the universe is a blob!
I have to wonder how many of the comments on here were written by a human. Certain ones, like Alex, sound to me more like a GPT-2 output.
@KevinLandry
So you’re saying it’s one big cat toy 🤔
I knew cats were aliens…it all makes sense now