The N.D. Zelinsky Institute of Organic Chemistry (IOC RAS) has developed a neural network that determines the molecular structure of substances from photographs taken with an optical or electron microscope, the Ministry of Education and Science reported.
Popular methods for studying molecular structures (NMR spectroscopy, mass spectrometry, X-ray diffraction) require the participation of qualified personnel and expensive equipment. The new approach reduces analysis time and makes it possible to use more affordable equipment.
To simplify the research technology, the authors proposed using machine learning. Specialists selected the structure of quaternary phosphonium salts (used in oil production, medicine, and other fields) to evaluate the performance of the neural network.
Researchers have proven that using machine learning to analyze images of chemical compounds is not just an experimental idea, but a real tool that can change many industries. This method will reduce the costs and time of analysis, which is especially important in the context of the rapid development of science and industry.
The developers plan to improve the accuracy of the neural network and include other chemical compounds in the analysis.
Earlier www1.ru reported that artificial intelligence was trained to quickly read data from utility meters.
Read materials on the topic:
MIPT taught artificial intelligence to write prescriptions and sick leave certificates
The Ministry of Digital Development has launched testing of generative AI on "Gosuslugi"