Observational bayesian optimization of quantum-computing operations
A method for calibrating a quantum-computing operation comprises: (a) providing a trial control-parameter value to the quantum computer; (b) receiving from the quantum computer a result of a characterization experiment enacted according to the trial control-parameter value; (c) computing a decoder estimate of an objective function evaluated at the trial control-parameter value based on decoding the result of the characterization experiment; (d) consuming the trial control-parameter value and the decoder estimate in a machine trained to return a model estimate of the objective function evaluated at the trial control-parameter value; and (e) selecting a new trial control-parameter value based on the model estimate.
1. A method for calibrating a quantum-computing operation on a quantum computer, the method comprising:
providing a trial control-parameter value to the quantum computer;
receiving from the quantum computer a result of a characterization experiment enacted according to the trial control-parameter value;
computing a decoder estimate of an objective function evaluated at the trial control-parameter value based on decoding the result of the characterization experiment;
consuming the trial control-parameter value and the decoder estimate in a machine trained to return a model estimate of the objective function evaluated at the trial control-parameter value; and
selecting a new trial control-parameter value based on the model estimate,
wherein said providing, receiving, computing, consuming, and selecting are enacted iteratively such that the trial control-parameter value converges to a control-parameter value that optimizes the objective function, and
wherein each estimation of the objective function contributes to an evolving model that serves as a surrogate for actual evaluation of the objective function.
2. The method of claim 1 further comprising furnishing a forward uncertainty of the decoder estimate based on decoding the result of the characterization experiment, wherein the machine is trained to return the model estimate further based on the forward uncertainty.
3. The method of claim 1 wherein the machine is further trained to return a backward uncertainty of the model estimate, wherein the decoder estimate is further based on the backward uncertainty estimate.
4. The method of claim 1 wherein the trial control-parameter value is one of a plurality of control-parameter values optimized in parallel.
5. The method of claim 1 wherein the result comprises a quantum-mechanical basis state.
6. The method of claim 1 wherein the result is one of a plurality of results, and wherein returning the decoder estimate comprises statistically estimating a function based on the plurality of results.
7. The method of claim 1 wherein the result comprises a real-valued function, and where returning the decoder estimate comprises transformation of the real-valued function.
8. The method of claim 1 wherein the result comprises a set of eigenvalues.
9. The method of claim 1 wherein the trial control-parameter value is a value of a hardware-control parameter.
10. The method of claim 1 wherein the trial control-parameter value is a value of a software-control parameter.
11. The method of claim 1 wherein the machine is trained via machine learning.
12. A quantum computer comprising:
a qubit register;
a qubit interface operatively coupled to the qubit register and configured to:
iteratively enact a characterization experiment on the qubit register according to a trial control-parameter value selected based on a model estimate of an objective function evaluated at the trial control parameter value, and
iteratively measure a quantum state of the qubit register pursuant to the characterization experiment to obtain a result of the characterization experiment enacted according to the trial control-parameter value; and
an output interface operatively coupled to the qubit interface and configured to output the result of the characterization experiment, wherein the objective function is iteratively estimated at each trial control-parameter value based on decoding the result of the characterization experiment using Bayes' rule, to give a decoder estimate,
wherein each estimation of the objective function contributes to an evolving model that serves as a surrogate for actual evaluation of the objective function, and wherein the trial control-parameter value and the decoder estimate are consumed in a machine trained to return the model estimate.
13. The quantum computer of claim 12 wherein the result comprises a quantum-mechanical basis state.
14. The quantum computer of claim 13 wherein the trial control-parameter value is a value of a hardware-control parameter.
15. The quantum computer of claim 12 wherein a forward uncertainty of the decoder estimate is furnished based on decoding the result of the characterization experiment, wherein the machine is trained to return the model estimate further based on the forward uncertainty and further trained to return a backward uncertainty of the model estimate, and wherein the decoder estimate is further based on the backward uncertainty.
16. A computer system comprising:
an acquisition engine configured to iteratively provide a trial control-parameter value to a quantum computer;
an input engine configured to iteratively receive from the quantum computer a result of a characterization experiment enacted according to the trial control-parameter value;
an observational decoder configured to iteratively compute a decoder estimate of an objective function evaluated at the trial control-parameter value based on decoding the result of the characterization experiment, wherein estimation of the objective function contributes to an evolving model that serves as a surrogate for actual evaluation of the objective function;
a trained machine configured to iteratively consume the trial control-parameter value and the decoder estimate, and to return a model estimate of the objective function evaluated at the trial control-parameter value,
wherein the acquisition engine is further configured to iteratively select a new trial control-parameter value based on the model estimate, such that the new trial control-parameter value converges to a control-parameter value that optimizes the objective function.
17. The computer system of claim 16 wherein the observational decoder is further configured to furnish a forward uncertainty of the decoder estimate based on decoding the result of the characterization experiment, and wherein the machine is further trained to return the model estimate further based on the forward uncertainty.
18. The computer system of claim 16 wherein the machine is further trained to return a backward uncertainty of the model estimate, and wherein the observational decoder is further configured to return the decoder estimate further based on the backward uncertainty.
19. The computer system of claim 16 wherein the acquisition engine, the input engine, the observational decoder, and the machine are configured for iterative execution, such that the trial control-parameter value converges to a control-parameter value that optimizes the objective function.