A group of Russian scientists from the Moscow Institute of Physics and Technology (MIPT or Phystech), together with American colleagues, have presented a new decentralized optimization algorithm for machine learning. It functions without a central server.
Unlike traditional methods, where it is necessary to predefine the parameters of the problem and the network topology, the MIPT scientists' algorithm uses local information. The researchers note that this approach significantly speeds up calculations, making the system more efficient and flexible.
The creation and optimization of new decentralized machine learning algorithms represents a significant step towards the development of more efficient and scalable machine learning systems in distributed environments.
As a reminder, machine learning is a branch of artificial intelligence focused on creating algorithms and models that allow computers to learn from data, as well as make predictions or make decisions without the need for explicit programming—a process in which the developer clearly and in detail sets out the instructions and rules that the computer must follow to solve a specific problem.
Read more on the topic:
Special molds for growing knee joint cells developed at MIPT
New possibilities for controlling superconductors discovered at MIPT
AI service for converting piano music into sheet music developed at MIPT
