Higher-order quantum algorithms for “generative” quantum learning

Mio Murao^1*

¹Department of Physics, The University of Tokyo, Tokyo, Japan

* Presenter:Mio Murao, email:murao@phys.s.u-tokyo.ac.jp

Efficiently learning the properties of quantum objects is one of the key anticipated applications of quantum computers. We develop quantum algorithms for “generative” quantum learning based on higher-order quantum computation, which transform a black-box quantum channel into either another quantum channel or a numerical quantity characterizing its properties. In this framework, a quantum computer acquires “quantum data” about the channel through black-box queries and directly produces the target channel or evaluates its properties internally, without requiring a full classical description of the black-box channel. We present two such “fully quantum” algorithms for generative quantum learning of black-box quantum channels: one for universal isometry adjointation and another for learning the singular value moments of unknown quantum channels.

Keywords: quantum learning, quantum algorithm, higher-order quantum computation