pianalytix machine learning

AI offers huge potential to augment current and future space exploration

Could the exact same computer algorithms which teach autonomous cars to drive safely assist identify nearby asteroids or find life in the universe? NASA scientists are attempting to figure out that by partnering with pioneers in artificial intelligence (AI) — companies such as Intel, IBM and Google — to employ sophisticated computer algorithms to problems in science.


Machine learning is a type of AI. It clarifies the most frequently used algorithms and other resources that allow computers to learn from data so as to make forecasts and categorize objects much faster and more precisely than a human being may. Thus, machine learning is broadly utilized to assist technology businesses recognize faces in photos or forecast what pictures people would enjoy. However, some scientists see software far beyond Earth.


“These technologies are extremely important, particularly for big data sets and particularly in the exoplanet field,” Arney says. “Since the data we are likely to get from future observations is going to be sparse and noisy. It is going to be really hard to understand. So using such tools has so much potential to help us.”


The four-year-old application is a partnership between the SETI Institute and NASA’s Ames Research Center, both based in Silicon Valley where startup-hatching incubators that bring talented people together to accelerate the development of breakthrough technologies are plentiful.


In NASA’s variant, FDL pairs science and computer technology early-career doctoral students with experts from the area agency, academia, and a number of the world’s biggest technology companies. Partner companies provide various combinations of hardware, algorithms, super-compute tools, financing, facilities and subject-matter specialists. All the AI methods developed at FDL will be publicly available, with a few currently helping identify asteroids, find planets, and forecast extreme solar power occasions.


“FDL feels like some really good musicians with various instruments getting together for a jam session at the garage, discovering something really cool, and saying,’Hey we have a ring here,”’ says Shawn Domagal-Goldman, a NASA Goddard astrobiologist who, with Arney, mentored an FDL group in 2018. Their staff developed a machine learning technique for scientists who plan to study the atmospheres of exoplanets, or planets beyond our solar system.


Since thousands of exoplanets have been discovered so far, making quick decisions about which ones have the most promising chemistry associated with habitability could help winnow down the candidates to only several that deserve further, and costly, investigation.

 An animated representation of all the multi-planet systems discovered in the Milky Way galaxy by NASA’s Kepler Space Telescope as of Oct. 30, 2018. The systems are shown together at the same scale as our Solar System (dashed lines).

Credits: Ethan Kruse/NASA Goddard


To this end, the FDL team Arney and Domagal-Goldman helped counsel, with technical support in Google Cloud, deployed a method called a”neural network” This technology may resolve super complicated issues in a process similar to the workings of the brain. In a neural network, billions of”neurons,” which are neural cells in the brain that help us form memories and make conclusions, connect with countless other people to process and transmit information. Another researcher not correlated with FDL had already used this latter system to analyze the atmosphere of WASP-12b, an exoplanet found in 2008, based on mountains of data collected by NASA’s Hubble Space Telescope. Could the Bayesian neural network do much better, the group wondered?


“We found out immediately that the neural network had improved precision than random forest in identifying the prosperity of various molecules in WASP-12b’s atmosphere,” Cobb says.


But besides better precision, the Bayesian technique offered something equally critical: it might tell the scientists certain it was about its prediction. “In areas where the data were not good enough to provide a very accurate result, this model was better at understanding that it wasn’t sure of the answer, which is actually important if we are to trust these predictions,” Domagal-Goldman says.


While the technique developed by this team is still in development, other FDL technology have been adopted in the real world. From 2017, FDL participants developed a machine learning program that could quickly create 3D models of nearby asteroids, accurately estimating their shapes, dimensions, and spin speeds. This information is vital to NASA’s attempts to detect and deflect threatening asteroids from Earth.


Traditionally, astronomers use easy computer applications to develop 3D models. The software assesses many radar dimensions of a moving asteroid and then helps scientists resisted its own physical properties based on changes in the radar signal.


“A skillful astronomer with regular compute resources, could form one asteroid in a few months,” says Bill Diamond, SETI’s president and chief executive officer. “So the question for the study team was: Could we speed it up?”

pianalytix machine learning
Eros is pitted with over 100,000 craters, the result of billions of years of impacts.
A 3D model of asteroid Eros.

Credits: NASA’s Scientific Visualization Studio


The answer was yes. The team, which included students from France, South Africa and the USA, also mentors from academia and by technology firm Nvidia, developed an algorithm that could render an asteroid at no more than four days.


The asteroid modeling, along with exoplanetary atmosphere analysis, are a couple FDL examples which show the guarantee in implementing sophisticated algorithms to the volumes of information collected by NASA’s more than 100 missions.


As NASA heliophysicist Madhulika (Lika) Guhathakurta notes, the space agency gathers about 2 gigabytes of information (and growing) every 15 minutes from its fleet of spacecraft. “But we examine just a fraction of the data, because we’ve limited people, resources and time. That is why we need to use these tools more,” she states.

An image of the Sun captured by NASA’s Solar Dynamics Observatory on Oct. 27, 2014. It shows a large active region (bottom right) erupting in a flare.


A lead on assignments focused on understanding and predicting the Sun’s effects on Earth, technology and astronauts in space, Guhathakurta has been with FDL for the previous 3 decades and has been a key architect in shaping this program.


Back in 2014, just four years after the assignment started, a detector stopped returning information related to extreme ultraviolet (EUV) radiation levels — information that contrasts with a ballooning of the Earth’s outer atmosphere and so affects the wellbeing of satellites, such as the International Space Station. Their computer program could do it by assessing data from other SDO devices, together with old information accumulated by the broken sensor through the four years it had been working, to infer what EUV radiation amounts that sensor would have discovered dependent on what the other SDO tools were observing at any given time. “We generated, essentially, a virtual sensor,” Guhathakurta says.


The potential of the sort of the tool is not lost on anyone. SETI head, Diamond, imagines a future in which these digital tools are integrated on spacecraft, a practice that will allow for lighter, less complicated and therefore more affordable missions. Domagal-Goldman and Arney envisage future exoplanet missions where AI technologies embedded on spacecraft are smart enough to make real-time science decisions, conserving the many hours necessary to communicate with scientists on Earth.


“However, these methods will not replace humans any time soon, because we will still need to check the results.” 


If you are Interested In Machine Learning You Can Check Machine Learning Internship Program

Also Check Other technical and Non Technical Internship Programs

Share on facebook
Share on whatsapp
Share on linkedin