Scientists from the Tinkoff Research laboratory have developed an innovative algorithm for training artificial intelligence (AI) and outperformed the existing reference method. Their method, called ReBRAC, turned out to be one and a half times more effective in tests on robotic simulators. This means that AI learns four times faster and with 40% better performance than existing algorithms.
Experts say ReBRAC solves the problem of AI being slow to adapt to new environments by ensuring it is better prepared for change. This algorithm allows AI to learn on the fly and adapt more successfully; This is particularly important when the robot moves from one surface to another, for example.
Tinkoff Research claims that such developments by Russian scientists help reduce the technological and digital divide between countries. Efficient algorithms require fewer computing resources, making it possible to create advanced technologies in countries with limited computing power.
At the international machine learning conference NeurIPS, scientists presented not only the ReBRAC algorithm, but also two open libraries in the field of offline reinforcement learning (Offline RL, ORL), which will facilitate the work of artificial intelligence specialists and allow them to avoid Reproduction of the results of scientific articles.
Source: Ferra
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
