Fountain Capacity - 06/04/2010

Lokesh Ravindranathan

Abstract: Point to point channel capacity was determined by Shannon towards the mid of last century. Many codes have been made to work at rates close to the capacity. One of the applications of using channels is to broadcast information to many devices simultaneously. Capacity of traditional codes are upper bounded to the capacity of the worst channel amongst them. Modern approach is to use fountain codes and achieve capacity close to the Shannon's capacity for individual channels.

The talk would also discuss on how rateless codes work in case of erasure channels. Since the codes are rateless, the capacity for fountain codes are defined from the perspective of decoder i.e ratio of information symbols transmitted to channel symbols received. The paper discusses on how fountain capacity is equal to Shannon's capacity for the memoryless channels. However, for memory based channels, the fountain capacity is lesser than the Shannon's capacity for channels.

Discussion would involve on how the fountain capacity will vary with feedback. Is there a change and if so, how will it change?