NCFU researchers are going to make AI systems both faster and less demanding, at the same time getting mobile with no loss in capacity.

Intelligence systems for data processing, such as speech or image detection, follow the principle on which neuron networks are built, and the precision is due to complicated calculations. This means that there is a need for a lot of resources, which, in turn, limits the potential for using AI.  Scientists all over the world have been trying to improve this data processing through hardware, whereas NCFU experts have taken a different path – improved calculation algorithms.

– We break large data sets into parts that are processing in independent parallels, which really facilitates the entire process, – Pavel Lyakhov (Head of Project; Associate Professor, Dept of Applied Mathematics & Mathematical Modelling) was quoted as saying. – Another advantage this method has to offer is that it allows reducing the resources demands, which means the entire chip can be smaller in size, which will inevitably reduce the cost and the energy consumption.

The research is held under a grant from the Russian Foundation for Basic Research and Russian President’s grants.

For more details, please see the e-version of the Neurocomputing journal.