Efficient data coding methods based on neural networks

Aleksandr Berezkin, Dmitry Kukunin, Alexey Slepnev, Ruslan Kirichek
This article discusses a neural network-based compression algorithm using noise-correcting codes. The use of this algorithm has a number of advantages, on the one hand, the noise-resistant code allows to get rid of potentially high overheads provoked by the use of a neural network, lowering the required value of model accuracy to the value determined by the correctability of the used code. On the other hand, a trained neural network allows data compression without prior transformations.