Table 1: The optimal parameters for Super Token Attention (STA) blocks in the each stage of Encoder/Decoder.Encoder \Decoder No. of layers Token size No. of Heads Stage - 1 1 16 x 16 2 Stage - 2 2 8 x 8 4 Stage
convolutional codesturbo code decoderredundant likelihood information updatingdecoding algorithmchannel valuessoft-input soft-output encodersThe paper presents a novel decoding algorithm for turbo codes, in which the likelihood values for redundant parts are updated in order that those values become more ...