Deep Learning Makes it Possible to Predict Life on Other Planets
Overview
- Researchers developed an artificial neural network that predicts the possibility of life on other planets
- The ANN can currently classify 5 planets
- A new metric called “probability-of-life” has been created to study this
Introduction
The topic of life on other planets, or extraterrestrial life, has always intrigued us. For centuries, scientists and researchers have been searching for answers but have always come up short due to lack of technological help. We all love those movies where we discover other planets habitable for human life and move away from Earth. These fictional stories are inching closer to becoming a reality.
Thanks to the advancements in deep learning, a group of researchers from the Centre for Robotics and Neural Systems (CRNS) at Plymouth University have developed a deep neural network that can analyse life patterns and identify how habitable a planet is.
The researchers trained artificial neural networks (ANNs) that can classify five types of planets:
- The early stages of Earth
- Earth as it is today
- Mars
- Venus
- Titan (Saturn’s moon)
ANNs are able to learn continuously through pattern recognition so the team fed them data using NASA’s Planetary Spectrum Generator (PSG). The data included hundreds of unique high resolution spectrum profiles of each planet mentioned above. Each profile had tons of parameters that defined the habitable nature of the planet.
The neural network then attempted to classify each of these planets. The results, so far, have been promising according to the project supervisor, Angelo Cangelosi. When the model was presented with a completely new profile, it adapted and performed well.
Our outlook on life is limited by human elements so the research team came up with a metric called “probability-of-life”. It is basically based on factors like the atmosphere and the orbit of the five planets/worlds it was tested on.
Our take on this
This model can be used for categorizing the different types of exoplanets. As a data scientist interested in this field, you can imagine how this technique can be used to identify prospective planets (other than the 5 mentioned above) for signs of life.
The findings and details about the ANN will be presented by this team today at the European Week of Astronomy and Space Science (EWASS) in Liverpool. We will update this article as and when we get more details about how accurate the final model has been so far. Watch this space!
Subscribe to AVBytes here to get regular data science, machine learning and AI updates in your inbox!
Senior Editor at Analytics Vidhya. Data visualization practitioner who loves reading and delving deeper into the data science and machine learning arts. Always looking for new ways to improve processes using ML and AI.
