Linear complexity universal decoding with exponential error probability decay

In this manuscript we consider linear complexity binary linear block encoders and decoders that operate universally with exponential error probability decay. Such scenarios may be relevant in wireless scenarios where probability distributions may not be fully characterized due to the dynamic nature of wireless environments. More specifically, we consider the setting of fixed length-to-fixed length near-lossless data compression of a memoryless binary source of unknown probability distribution as well as the dual setting of communicating on a binary symmetric channel (BSC) with unknown crossover probability. We introduce a new 'min-max distance' metric, analogous to minimum distance, that addresses the universal binary setting and has the same properties as that of minimum distance on BSCs with known crossover probability. The code construction and decoding algorithm are universal extensions of the 'expander codes' framework of Barg and Zemor and have identical complexity and exponential error probability performance.