A research group led by Associate Professor Teruyasu Mizoguchi of the Institute of Industrial Science, University of Tokyo has succeeded in predicting the structure of a complex interface with simple artificial intelligence created by machine learning.The calculation that takes 22 years with the conventional method can be completed in about 3 hours, and it seems that the development speed of excellent substances will be accelerated by using this method.
According to the University of Tokyo, the interface refers to the interface where two phases such as gas and liquid, liquid and solid, and solids are in contact with each other, but the structure is extremely complicated and requires a huge amount of calculation. Until now, it has been difficult to predict.
The research group of Associate Professor Mizoguchi and others challenged prediction by using a method called virtual screening in machine learning that finds the laws hidden in the data by statistically processing the data.In virtual screening, when a computer learns data, it creates a simple artificial intelligence device called a regression device.
By passing this, it became possible to predict the interface without a huge amount of calculation.Compared to the conventional method, the calculation speed is actually 6 times faster.
Interfaces are deeply related to the functions of various substances such as batteries and catalysts.For this reason, this research will be adopted as a research project of the Strategic Creative Research Promotion Project of the Japan Science and Technology Agency, and will continue with its support.The research results so far have been published in the scientific journal "Science Advance" published by the American Society for the Promotion of Science.
