Binary classification with classical instances and quantum labels

摘要: In classical statistical learning theory, one of the most well-studied problems is that binary classification. The information-theoretic sample complexity this task tightly characterized by Vapnik-Chervonenkis (VC) dimension. A quantum analog task, with training data given as a state has also been intensely studied and now known to have same its counterpart. We propose novel version classification considering maps input output corresponding classical-quantum data. discuss strategies for agnostic realizable case study their performance obtain upper bounds. Moreover, we provide lower bounds which show our are essentially tight pure states. particular, see in w.r.t. dependence on accuracy, confidence VC-dimension.