Training method and device based on improved protograph neural decoder

A training method based on an improved protograph neural decoder, comprising:

• constructing a to-be-trained decoding network based on an initial variable sub-network layer, an initial check sub-network layer and a preset shuffled belief-propagation (BP) sub-network layer, wherein the to-be-trained decoding network comprises a plurality of hidden layers and an input convolutional layer;
• and the to-be-trained decoding network is constructed through steps of:
• successively connecting the initial variable sub-network layer, the initial check sub-network layer and the preset shuffled BP sub-network layer to obtain a hidden network group;
• and
• successively connecting a plurality of hidden network groups to form a hidden layer;
• and successively connecting the plurality of hidden layers after the input convolutional layer to obtain the to-be-trained decoding network;
• and
• updating and training the initial variable sub-network layer, the initial check sub-network layer and the preset shuffled BP sub-network layer by calculating log-likelihood ratio (LLR) based on a preset mean square error loss function and a preset decoder objective function to obtain a target protograph neural decoder;
• wherein the preset mean square error loss function is configured to calculate a loss value between output information of the check sub-network layer and output information of the preset shuffled BP sub-network layer;
• and
• the target protograph neural decoder comprises an optimized variable sub-network layer, an optimized check sub-network layer and an optimized shuffled BP sub-network layer.