We find “combinatorially optimal” (guaranteed by the degree-sequence alone) independent sets for graphs and hypergraps in linear space in the semi-streaming model. We also propose a new output-efficient streaming model, that is more restrictive than semi-streaming (n·logO(1) n space) but more flexible than classic streaming (logO(1) n space). The algorithms in this model work in poly-logarithmic space, like in the case of the classical streaming model, but they can access and update the output buffer, treating it as an extra piece of memory. Our results form the first treatment of the classic IS problem in the streaming setting.
Bjarni V. Halldórsson, Magnús M. Hal