“We are currently interested in these ANNs because they may enable a hypothetical, intelligent starship capable of scanning an exoplanet in range,” Christopher Bishop, a researcher at the University of Plymouth’s Robotics and Nervous System Research Centre, said in a statement. He added, “We are also looking at using a large deployable planar Fresnel antenna that could transmit data from interstellar probes to Earth over long distances. The application of this technology is essential for future robotic spacecraft.”

The Plymouth team trained their neural network to classify planets into one of these five distinct categories based on how similar the planets to be predicted would be to present day Earth, early Earth, Mars, Venus and Titan, Saturn’s largest moon.

These planets are similar in that they are rocky and have atmospheres, although their composition is often different. According to the new evidence, Mars and Titan are the two most promising targets for extraterrestrial life in the solar system after Earth, and “Saturn’s largest moon Titan is one of the most likely places in the solar system to harbor extraterrestrial life,” said Kevin M.Gill of NASA’s JPL Space Sciences. For scorched Venus, there are also some possible conditions for life.



This is an infrared image of Saturn’s moon Titan from NASA’s Cassini spacecraft. The measurements indicate that Titan has the highest habitable class other than Earth, based on factors such as energy availability, surface conditions and atmospheric characteristics. — NASA/JPL/ University of Arizona/University of Idaho

ANN is a system that tries to mimic the way the human brain learns. As one of the most commonly used tools in machine learning, it works well when it comes to making complex pattern recognition of large data sets, a process that is difficult and time-consuming for human scientists. So an ANN that can make predictions about the habitability of a large number of planets could save scientists time and help them focus on the most promising targets.

The scientists fed the ANN model with atmospheric observations from five types of planets: spectral values, which are classified using “probability of life” as a measurement criterion since life is currently known only to exist on Earth. The criteria are based on the atmospheric and orbital properties of the five planets.

Bishop has trained the network to determine its ability to harbor life with hundreds of different spectral profiles, each with hundreds of parameters used to measure the habitability of planets. So far, the network has performed well on a test spectrum that has never been seen before.



The ANN input is from the spectral values of the test planets, and the output layer is based on the classification results of the “probability of life”, which is measured by the similarity of the input data to the five solar system planets. The input data, after being feature-mapped through a series of interconnected hidden layers in the network, is able to “learn” specific planet types that correspond to spectral lines. — Bishop/ Lemouth University

“Given the results so far, this method may be able to classify different types of exoplanets very efficiently by using the classification results of different types of exoplanets from Earth and near-Earth observatories,” project director Angelo Cangelosi said in a statement.

The researchers hope their technology will come into play when NASA’s James Webb Space Telescope and the European Space Agency’s Ariel space mission begin to provide more detailed observations of the atmospheres of potentially habitable exoplanets.


The original article was published on April 6, 2018

Author: Xiao Pan

This article is from xinzhiyuan, a partner of the cloud community. For relevant information, you can follow the wechat public account “AI_era”

ANN gives aliens nowhere to hide? Ai predicts signs of life on other planets