Wide applicability
To train a neural network at the declared speed, less sophisticated equipment is required compared to classical neural networks. For example, instead of a desktop – a laptop, instead of a workstation – a desktop, instead of a server – a workstation, instead of a supercomputer – a server. The network is compatible with different hardware: from microchips, video cards, and phones to supercomputers, and boosts their computing capabilities by a multiple.
Security and data protection
Calculations can be performed at client’s own facilities to guarantee the safety and confidentiality of data. Built-in high security and reliability.