Locally decodable codes

Abstract

Locally Decodable Codes (LDCs) are a special kind of error-correcting codes. Error-correcting codes are used to ensure reliable transmission of information over noisy channels as well as to ensure reliable storage of information on a medium that may be partially corrupted over time (or whose reading device is subject to errors). In both of these applications the message is typically partitioned into small blocks and then each block is encoded separately Such encoding strategy allows efficient random-access retrieval of the information, since one needs to decode only the portion of data one is interested in. Unfortunately, this strategy yields very poor noise resilience, since in case even a single block (out of possibly tens of thousands) is completely corrupted some information is lost. In view of this limitation it would seem preferable to encode the whole message into a single codeword of an error-correcting code. Such solution clearly improves the robustness to noise, but is also hardly satisfactory, since one now needs to look at the whole codeword in order to recover any particular bit of the message (at least in the case when classical error-correcting codes are used). Such decoding complexity is prohibitive for modern massive data-sets. © 2011 Springer-Verlag Berlin Heidelberg.