With the help of machine learning technology, Japanese scientists can predict in detail the possible impact of the tsunami in less than one second, while the traditional method takes 30 minutes, which can win valuable time for people to take appropriate action. The relevant research was published in the latest issue of Nature Communication. The researchers pointed out that the disastrous tsunami that hit northeastern Japan on March 11, 2011 killed about 18500 people. If the early warning of the tsunami can predict the water level at different locations along the coastline and inland more quickly and accurately, many lost lives may be saved. The northeastern coast of Japan now has the world's largest sensor network for monitoring undersea movements. The 150 offshore stations that make up the network will provide tsunami early warning. However, in order for these early warnings to be meaningful, the data obtained by the sensors need to be converted into tsunami height and other values. To do this, scientists usually need to use standard computers to solve difficult nonlinear equations, which generally takes about 30 minutes to get results. In the latest research, the Japanese Institute of Physics and Chemistry predicted that the scientific laboratory would use machine learning technology to shorten the calculation time to less than 1 second. Since tsunamis rarely occur, the team used more than 3000 computer-generated tsunami events to train the machine learning system, and then tested 480 simulated tsunami scenarios and 3 actual tsunami scenarios. The results show that the model based on machine learning can achieve the same accuracy with only 1% of the previous calculation workload. However, this latest method is only applicable to large tsunamis higher than 1.5 meters. The research team is improving the relevant method to make it applicable to smaller tsunamis. The researchers said that the same machine learning method can also be used in other disaster scenarios. (Reporter Liu Xia) (News Agency)