New efficient algorithms for compressing biased input sequences into accurate summaries of a target distribution, achieving better-than-i.i.d. error rates.
