ABSTRACT
A common question arising in a wide variety of practical applications is how to infer an unobserved high-dimensional “state of the world” from a limited number of observations. Examples include finding a subset of genes responsible for a disease, localizing brain areas associated with a mental state, diagnosing performance bottlenecks in a large-scale distributed computer system, reconstructing high-quality images from a compressed set of measurements, and, more generally, decoding any kind of signal from its noisy encoding, or estimating model parameters in a high-dimensional but small-sample statistical setting.
