DeepMind Technologies Limited, together with British meteorologists, presented a new development: the artificial intelligence DeepMind, which is able to determine the probability of precipitation in the next hour and a half.
According to world standards, forecasters make up a weather forecast based on a whole data system. This is information from hydrological and agrometeorological stations, analysis of a number of changes on land, air and water. However, forecasters are not always able to determine accurate forecasts for the next 2 hours.
To make predictions as accurate as possible, the developers of DeepMind used DGMs - deep generative mechanisms. They analyze the received data, form samples and create a set of predictions based on the ensemble method.
At the same time, DeepMind takes into account the previous radar characteristics using two generative neural networks. This allows the AI to "capture" both temporal and spatial dimensions, preventing the likelihood of erroneous predictions.
This technique resembles the work of neural networks that create video. In fact, it is a chain of frames in a specific sequence.
AI from DeepMind is able to predict weather conditions on a small scale, which, due to their natural features, are almost impossible to predict with a conventional meta. The system was evaluated by more than 50 experts. They all confirmed that the accuracy of the weather forecast from DeepMind is 99%.
At home, standard thermometers are suitable for measuring temperature. You can buy them in
this section.