Earth 2.0 Discovered with the Aid of an Innovative Deep Learning Algorithm

How useful is machine learning for astronomers while searching for Earth-like exoplanets?This is what a recently accepted study to Astronomy & Astrophysics aims at as a team of international researchers studied how a novel neural network based algorithm can be employed for the detection of earth like exoplanets using RV detection method. This study could assist astronomers to design better approaches that are used in distinguishing Earth-like exoplanets from RV data where the host star gives extreme activity.

The study notes, “Machine learning is one of the most efficient and successful tools to handle large amounts of data in the scientific field. Many algorithms based on machine learning have been proposed to mitigate stellar activity to better detect low-mass and/or long period planets. These algorithms can be classified into two categories: There are two types of machine learning, which are known as supervised learning and unsupervised learning. The benefit behind supervised learning is that the proposed model includes a great number of variables and the ability to generate relatively precise predictions of the course, according to the training data.

The researchers used their algorithm for three stars to test it for the effectiveness for finding exoplanets within the stellar activity data, including the sun, Alpha Centauri B (HDTau ceti is approximately 12 light-years from Earth and IC 348 is located approximately 3 light-years from Earth. After the researchers input simulated planetary signals in their algorithm, they were able to retrieve simulated exoplanets with possible orbital periods between ten to 550 days for the sun, ten to 300 days for Alpha Centauri B and ten to 350 days for Tau ceti.

It’s important to note that Alpha Centauri B currently has had several potential exoplanet detections but non confirmed while Tau ceti currently has eight exoplanets listed as “unconfirmed” within its system.

Moreover, the algorithm generated these outputs: Alpha Centauri B and Tau ceti both may have exoplants four times the size of the earth and in the stars’ habitable regions. By adding more information about stellar activities into the algorithm, the researchers found that the algorithm is actually capable of identifying a simulated exoplanet at least 2. It was two times the size of the Earth and it orbited at the distance of the Earth from the Sun.

The study concluded in the paper when it stated, “In this paper, we provided a neural network approach that could effectively suppress stellar activity at the spectrum level to improve the detection on low mass planets within the period range of a few days to a few hundred days which belonged to the habitable zone for the solar-type stars. ”

According to the study, RV data can be used to identify Earth-like exoplanets, but they added that other data such as transit time, phase and space-based photometry can also be employed for the same purpose. They stress that this could be done by the European Space Agency’s PLATO space telescope mission that is under construction and planned for launch at some time in 2026. At launch it will be situated at the Sun-Earth L2 Lagrange point, which is 1. 5 million km beyond the Earth and on the opposite side of the Sun from Earth where it will scan up to one million stars pursuing terrestrial (rocky) exoplanets using the transit method.

This comes as NASA has confirmed 5,632 exoplanets to date, out of which only 201 belong to the terrestrial category, which is further promising for the upcoming PLATO mission to detect several more terrestrial exoplanets within the Milky Way Galaxy.

In the following years and decades, how does machine learning support astronomers in finding more Earth-like exoplanets?The only way to find this out is to wait and this is why we do science and not witch craft.

Do not forget to share your opinion with us to provide you with the best posts !