It’s similar to how paleontologists use certain known fossils Indexed Fossil Until assessing rock formations and ancient environments so far, astronomers look for specific patterns of light emissions from space to mark the age of space history. For example, early galaxies give the UV rays that originate from electrons in hydrogen atoms to the type of ultraviolet rays that exert from the second lowest to the lowest energy state. Lyman Alpha or ly⍺ Emission.
For decades, astronomers have associated ly⍺ emissions with periods within billions of years of a big bang called the Big Bang. The era of reionizationwhen the average speed of star formation in galaxies was much higher than today. When they find a galaxy that emits light strongly, they classify it into ly⍺Emitter or Lae And we can be sure that it goes back to the era of reionization. Observing Laes, astronomers talk more about the history of the Milky Way and other galaxies like us.
However, researchers face confounding factors when looking for Laes. The expansion of the universe distorts light in a process called Cosmological redshift. However, more prominently Dustboth Intergalacticcovers the light. While astronomers can analyze the full light of light from the galaxy to find evidence of ly⍺ emissions, it would be much faster to develop tools to predict whether a galaxy is likely to be a LAE based on more readily available measurements.
One team of astronomers developed a model for this problem only Machine Learning A technique known as a Neural Networks. This technique replicates how neurons in the brain function, with several interconnected layers receiving and transmitting signals based on initial inputs and generating final outputs.. The trick is that the programmer knows what inputs to input and what output they expect in the end. The algorithm itself needs to know how best to set up a central connection, what to look for, and how to rank the importance of each input.
The team began with data from two surveys of light sources in space: 926 galaxies VanderOf these, only 520 are laes, starting from 507 Musethey were all laes. They trained the algorithm using 80% of this data to explicitly communicate which sources are actual LAES and which sources are not. They saved the remaining 20% of the data for testing.
Through this initial test, the team identified six parameters of neural networks to focus on evaluating galaxies for LAE potential. These parameters were the rate of star formation, total star mass, UV brightness, UV emission patterns, age, and dust. They programmed the network to output an estimate of the probability that a particular galaxy is a LAE, and thought that what was above 70% meant that the algorithm classified it as an LAE.
When we created a neural network using training data, the team tested several additional rounds. Using early test data, their networks found that they correctly identified the network in 77% of the time, as there was only a 14% chance of false positives. When they looked at what their network prioritized to make these predictions, they found that the most important factors were the galaxy’s UV emission pattern, its UV brightness, and the mass of its star.
Following this initial success, the team applied the network to another investigation. cosmos2020and a subset of that raise, SC4Kwith fewer details than the training data survey. From these datasets, the team’s neural network identified true Laes for 72% of the time.
The team’s final results came when they applied neural networks to data from NASA’s new telescope. jwst. The ultimate goal in their model is to study the distant past of the universe, and JWST aims to see better-looking sources than ever before, so the success of the test is Already checking the results of LAE from JWST It will be a good sign of future success. They found a true positive rate of 91% in JWST data, showing the validity of their approach and illuminated the path to know more about the history of the universe.
Post view: 1,100
Source: sciworthy.com
