BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Drupal iCal API//EN X-WR-CALNAME:Events items teaser BEGIN:VEVENT UID:632edb11a998a DTSTART;TZID=America/Toronto:20220818T140000 SEQUENCE:0 TRANSP:TRANSPARENT DTEND;TZID=America/Toronto:20220818T150000 SUMMARY:Tight bounds for Quantum Learning and Testing without Quantum Memor y CLASS:PUBLIC DESCRIPTION:Summary \n\nJERRY LI - MICROSOFT RESEARCH\n\nIn this talk\, we consider two fundamental tasks in quantum state\nestimation\, namely\, qu antum tomography and quantum state\ncertification. In the former\, we are given n copies of an unknown\nmixed state rho\, and the goal is to learn i t to good accuracy in trace\nnorm. In the latter\, the goal is to distingu ish if rho is equal to\nsome specified state\, or far from it. When we are allowed to perform\narbitrary (possibly entangled) measurements on our co pies\, then the\nexact sample complexity of these problems is well-underst ood. However\,\narbitrary measurements are expensive\, especially in terms of quantum\nmemory\, and impossible to perform on near-term devices. In l ight of\nthis\, a recent line of work has focused on understanding the\nco mplexity of these problems when the learner is restricted to making\nincoh erent (aka single-copy) measurements\, which can be performed much\nmore e fficiently\, and crucially\, capture the set of measurements that\ncan be be performed without quantum memory. However\, characterizing\nthe copy co mplexity of such algorithms has proven to be a challenging\ntask\, and clo sing this gap has been posed as an open question in\nvarious previous pape rs.\n DTSTAMP:20220924T102521Z END:VEVENT END:VCALENDAR