A machine learning method called deep learning, which is widely used in face recognition and other image and speech recognition applications, has shown that it can help astronomers analyze and understand images of galaxies and how they arise develop. 19659002] In a new study published in the Astrophysical Journal and available online, researchers used computer simulations of galaxy formation to train a deep learning algorithm that then performed surprisingly well in the analysis of Hubble Space Telescope galaxies.
The researchers used the results of the simulations to generate pseudo-images of simulated galaxies, as they would look in observations of the Hubble Space Telescope. The dummy images were used to train the deep learning system to identify three key phases of galaxy evolution previously identified in the simulations. The researchers then gave the system a large amount of actual Hubble images for classification.
The results showed a remarkable consistency in the classifications of the neural network of simulated and real galaxies.
"We did not expect anything to happen" I am amazed at how powerful this is, "said co-author Joel Primack, Professor Emeritus of Physics and member of the Santa Cruz Institute for Particle Physics (SCIPP) at UC Santa Cruz. "We know that the simulations have limitations, so we do not want to argue too much. But we do not believe that this is a lucky one. "
Galaxies are complex phenomena that change their appearance as they evolve billions of years, and images of galaxies can only deliver snapshots in time Looking at the Universe and seeing "in time" earlier galaxies (because of the time it takes light to cover cosmic distances), but tracking the evolution of a single galaxy over time, is only possible in simulations: the comparison of simulated galaxies with observed galaxies can reveal important details of the actual galaxies and their probable history.
In the new study, researchers were particularly interested in a phenomenon observed in the simulations at the beginning of the evolution of gas-rich galaxies, large gas flows into the center of a galaxy form a small, dense, star-forming region called "Blue Nugget "is called. (Young, hot stars emit short "blue" wavelengths of light, so blue indicates a galaxy with active star formation, while older, colder stars emit more "red" light.)
In simulated and observed data, the computer program found that the "Blue Nugget" phase occurs only in galaxies with masses within a certain range. Subsequently, star formation in the central region is quenched, resulting in a compact "red nugget" phase. The consistency of the mass domain has been an exciting finding as it suggests that the deep learning algorithm itself identifies a pattern that results from an important physical process in real galaxies.
"Maybe in a certain size range Galaxies have just the right mass for this physical process," said co-author David Koo, emeritus professor of astronomy and astrophysics at UC Santa Cruz.
The researchers used state-of-the-art galaxy simulations (the VELA simulations)) developed by Primack and an international team of collaborators, including Daniel Ceverino (University of Heidelberg), who led the simulations, and Avishai Dekel (Hebrew University), whose analysis and analysis Led interpretation and developed new physical concepts based on them. However, all of these simulations are limited in their ability to capture the complex physics of galaxy formation.
In particular, the simulations used in this study do not include feedback from active galactic nuclei (injection of energy from radiation when gas is accreted) through a central supermassive black hole). Many astronomers consider this process as an important factor for star formation in galaxies. Nevertheless, observations of distant, young galaxies seem to provide clues to the phenomenon leading to the blue nugget phase in the simulations.
For the observation data, the team used images of galaxies generated by the CANDELS project (Cosmic Assembly Near- (Deep Extragalactic Legacy Survey), the largest project in the history of the Hubble Space Telescope, first author Marc Huertas-Company Astronomers at the Paris Observatory and Paris Diderot University have already pioneered the use of Deep Learning methods on galaxy classifications with publicly available CANDELS data.
Koo, a CANDELS co-investigator, has invited Huertas. Companies are visiting UC Santa Cruz to continue this work, and Google has supported its work on deep learning in astronomy by donating research funding to Koo and Primack so that Huertas-Company can spend the last two summers in Santa Cruz with another visit in the summer of 201
"This project was just one of several ideas we had," Koo said. "We wanted choose a process that theorists can clearly define from the simulations, and that has something to do with what a galaxy looks like, and then the deep learning algorithm has to look in the observations for it. We are just beginning to explore this A new way to combine theory and observation. "
For years, Primack has been working closely with Koo and other astronomers at UC Santa Cruz to compare the simulation of galaxy formation and evolution with the CANDELS observations." The VELA simulations have been very successful in helping us Understand the CANDELS observations, "Primack said," nobody has perfect simulations. As we continue this work, we will continue to develop better simulations. "
Deep learning has the potential to expose aspects of observational data that people can not see, but the disadvantage is that the algorithm is like a" black box. " so it's hard to know which features in the data the machine uses to make its classifications, but network query techniques can identify which pixels in an image contributed most to classification, and researchers tested such a method on their network.
"Deep learning looks for patterns, and the machine can see patterns that are so complex that we humans do not see them," said Koo. "We want to test this approach a lot more, but in this proof-of-concept Study, the machine seemed to successfully find the various stages of galaxy evolution identified in the simulations in the data. "
I In the future, he said, astronomers will have much more observation data to analyze as a result of large surveying projects and new telescopes, such as the Great Synoptic Surveying Telescope, the James Webb Space Telescope, and the Wide-Field Infrared Survey Telescope. Deep Learning and other machine learning methods could be powerful tools to make those huge records meaningful.
"This is the beginning of a very exciting time for the use of modern artificial intelligence in astronomy," said Koo.
University of California – Santa Cruz
All about the robots on Earth and beyond!
We need your help. SpaceDaily's news network continues to grow, but earnings have never been so difficult to sustain.
With the advent of ad blockers and Facebook, our traditional sources of revenue through high-quality network advertising continue to decline. And unlike many other news sites, we do not have a paywall – with those annoying usernames and passwords.
Our reporting takes time and effort to publish 365 days a year.
If you find our news sites informative and useful, then consider becoming a regular supporter or make a one-off contribution now.
$ 5 Once Calculated
Credit Card or PayPal
SpaceDaily Monthly Supporter
$ 5 Calculated Annually
Researchers design "soft" robots that move alone
Houston TX (SPX) April 19, 2018
If Star Wars ™ R2-D2 is your idea of a robot, think again. Researchers led by a University of Houston engineer reported a new class of soft robots consisting of ultrathin sensors, activating electronics and temperature-sensitive artificial muscles that can adapt and crawl to the environment, similar to the movement of an inchworm a caterpillar.
Cunjiang Yu, Bill D. Cook Assistant Professor of Mechanical Engineering, said that potential applications range from surgery and rehabilitation to … Read more