Appears in collection : 2016 - T1 - WS3 - Central workshop

An adversary wishes to corrupt stored (or transmitted) data, but operates in an information-limited manner. Examples of such limitations may include only viewing some subset of the data or some noisy version of it, viewing those subsets causally, or with the encoder getting feedback about the corruption. We determine the capacity of some classes of such channels, and computationally efficient schemes achieving these capacities in some models (in particular over “large alphabets”). This is an overview of a long line of classical results, and also work done over the last few years (with an emphasis on a flurry of recent results) in collaboration Bikash Kumar Dey, Anand Dilip Sarwate, Michael Langberg, Zitan Chen, Mayank Bakshi, Qiaosheng Zhang (Eric), Alex Sprintson, and Swanand Kadhe.