Rateless codes, as proposed in Shulman's Ph.D thesis, can be used
for communication over unknown channels where the transmitter sends
a fixed amount of bits using codewords that can be as long as
desired, until the receiver (who may use universal decoder) gets
enough information, via the specific channel, to decide on the
transmitted bits. The application of rateless codes assume feedback
(whose rate can be as small as desired) to indicate that the message
was decoded. In this talk I will describe new findings related to
rateless codes for communication over unknown channels:
\begin{enumerate}
\item Rateless codes can be used for communication over ``individual channels'',
i.e., channels for which no mathematical model or any other
assumption is made, only the transmitted input sequence and the
resulting output sequence are known a-posteriori. Rateless codes are
used to achieve the desired rate given by the a-posteriori empirical
behavior of the input/output pair (joint work with Yuval Lomnitz)
\item Rateless codes can be used in multiuser environment and in
cases where side information at the transmitter and/or receiver that
is unknown a-priori, may be available. \end{enumerate}