The newly developed method will make it far simpler to predict the environmental impact of any substance, the study’s first author said
MOSCOW, July 20. /TASS/. An international research team, together with scientists from the Skolkovo Institute of Science and Technology (Skoltech), has created a neural network to predict complex properties of organic substances using a minimal set of input data, Skoltech's press office reported.
The results of the study were published in the Journal of Physics: Condensed Matter.
"Our newly developed method will make it far simpler to predict the environmental impact of any substance. Later on, our development can also be used to forecast the properties of exotic molecules and principally new compounds for which the existing tools of searching for the structure-property relationship do not work," the study’s first author, Skoltech PhD Sergey Sosnin stated.
The essence of the method is carrying out a physical-chemical calculation and using this data to define 3D densities of hydrogen and oxygen around a molecule under study. Then an analysis of the obtained information with 3D convolutional neural networks is conducted.
Sosnin also noted that the main achievement of their scientific work is the creation of "a universal way to describe a molecule for further transmission of its "image" to a 3D convolutional neural network." The convolutional neural network is a special architecture of artificial neural networks aimed at efficient image recognition.
Skoltech is a non-governmental scientific and educational institution established in 2011 under the participation of Massachusetts Institute of Technology.