Approaching Blokh-Zyablov Error Exponent with Linear-Time Encodable/Decodable Codes

Abstract: Guruswami and Indyk showed in [1] that Forney's error exponent can be achieved with linear coding complexity over binary symmetric channels. This paper extends this conclusion to general discrete-time memoryless channels and shows that Forney's and Blokh-Zyablov error exponents can be arbitrarily approached by one-level and multi-level concatenated codes with linear encoding/decoding complexity. The key result is a revision to Forney's general minimum distance decoding algorithm, which enables a low complexity integration of Guruswami-Indyk's outer codes into the concatenated coding schemes.