Specialists from Cognitive Pilot (a subsidiary of Sber and Cognitive Technologies) have developed an automatic mechanism to accelerate neural network training and correct errors during training.
Reducing training time and increasing its efficiency ultimately reduces the development time of AI autopilot systems by orders of magnitude, which ultimately allows users of our systems - agricultural holdings, passengers of smart trams, etc. - to constantly use the newest and most advanced solutions.
The new technology, called Cognitive Neural Network Hospital, reduces the training time of artificial intelligence (AI) by 15-20 times. In the first stage, it finds the weakest zones, and then identifies new data that is similar to perfect miscalculations, but does not completely coincide with them. This allows you to analyze and classify errors.
Earlier www1.ru reported that Cognitive Pilot created a navigation system for smart trams.
Read related materials:
A fully unmanned tram will appear in St. Petersburg in 2026
Unmanned tractors with the Cognitive Pilot system have learned to work without GPS
