The most popular and comprehensive Open Source ECM platform
Let us know more
Geometric Deep Learning: Recognizing Patterns on Curved Surfaces
Geometric Deep Learning (GDL) is a branch of artificial intelligence that focuses on learning patterns that occur on curved surfaces. It is particularly useful in helping biology and chemistry researchers unlock the interactions of molecules and proteins. The area of study was first discussed in a paper in 2016 by Michael Bronstein, now a professor at Imperial College in London.
Geometric Deep Learning has been used to study the interactions of proteins in the human body. GDL can be used to predict how proteins will interact and bind. The technique was used to study the virus that causes COVID-19. It has been used to study the cortical folding of the human brain.
Bruno Correia, assistant professor, said that “we are trying to see what fingerprints are in that virus [SARS-CoV-2/COVID-19]. It does seem like the virus has some places where we could try to attack it, besides the ones that we already knew. If we could design new proteins based on the surface fingerprints of the viral protein in order to inhibit the way the virus invades host cells, that would be pretty exciting.”
The 3D-recognition capabilities of GDL may have unexpected applications, for example, authentication. Researchers are taking fingerprint recognition to the next level by using the 3D characteristics of a hand. For example, the ridges, finger size, and distance between muscles. This could even be combined with gestures to capture motions unique to an individual. The recognition of all of these details together would create a virtually unhackable signature and, unlike facial recognition, the shape of the hand doesn’t include recognizable personal information, so there aren’t problems with privacy and discrimination.
It’s likely that Geometric Deep Learning could have useful applications in 3-dimensional design and search, for example, it may be useful for space optimization problems like pipe layout, and for complex fluid and aerodynamic design problems.
