In this paper, we study the problem of using statistical
query (SQ) to learn highly correlated boolean functions, namely, a
class of functions where any
pair agree on significantly more than a fraction 1/2 of the inputs.
We give a limit on how well ... more >>>