Benchmarking Algorithms Beyond the Horizon of Classical Computability

I will describe a new breakthrough capability for benchmarking the performance of quantum gates and entire algorithms run on large-scale but error prone quantum computing hardware. I will first introduce a scalable method... [ view full abstract ]

I will describe a new breakthrough capability for benchmarking the performance of quantum gates and entire algorithms run on large-scale but error prone quantum computing hardware. I will first introduce a scalable method for characterizing the error rates in universal gate sets acting on multi-qubit quantum information processors during a single clock cycle. This `cycle benchmarking' method offers a dramatic improvement over existing tools, such as interleaved randomized benchmarking, because it requires only single-qubit randomizing gates and thus can produce much more precise and fine-grained estimates of the error model, including the length-scale of spatial correlations and magnitude of cross-talk errors for arbitrary parallelized gates.  This method is practical to implement and has been implemented experimentally on quantum processors with up to 10 qubits for both parallelized local gates and multi-qubit entangling gates. I will then describe how cycle benchmarking enables a means of directly bounding the precision of the output under universal circuits on error-prone quantum hardware when implemented via randomized compiling. We call this a quantum (circuit) capacity assessment of performance (QCAP). This is an enabling new capability for validating hardware and certifying the performance of broad classes of algorithms that will soon be tested on near-term quantum devices that can compute into the regime that lies beyond the horizon of classical computability.