![Computational Complexity of Statistical Inference_hi-res logo](/sites/default/files/styles/workshop_banner_sm_1x/public/2023-03/Computational%20Complexity%20of%20Statistical%20Inference_hi-res.jpg?h=ccd3d9f7&itok=qCiIzWU5)
Abstract
We look at the computational problem of learning halfspaces under various noise models. In particular, we prove new lower bounds in the statistical query model for learning a halfspace to small error even when the optimal error is very small.