Neural Network Coding in Data Compression Systems in Communication Channels

The article describes the use of a neural network implementation of BCH codes encoding and decoding devices for error correction in on-the-fly data compression systems. The user payload vector is encoded in the BCH code with additional redundancy, which provides linear data separability and better convergence when training the autoencoder. On the receiving side, the decoding device operates in the error correction mode, which makes it possible not to impose strict requirements on the accuracy of data recovery when training the autoencoder. Examples of such devices are described in the article.