Using a quantum processor unit to preprocess data

1. A method comprising:

obtaining input data for a computer process configured to run on a computer processing unit;

by operation of a classical processor, computing randomized parameter values for variable parameters of a quantum logic circuit, wherein the randomized parameter values are computed from the input data and a set of random numbers;

causing a quantum processor to execute the quantum logic circuit having the randomized parameter values assigned to the variable parameters, wherein the execution of the quantum logic circuit produces quantum processor output data; and

providing data derived from the quantum processor output data to the computer processing unit configured to run the computer process.

2. The method of claim 1 , comprising obtaining the set of random numbers by sampling a probability distribution.

3. The method of claim 1 , wherein a parameterized quantum circuit comprises the quantum logic circuit having the randomized parameter values assigned to the variable parameters, and at least a portion of the parameterized quantum circuit corresponds to a unitary operator that depends on some or all of the randomized parameter values.

4. The method of claim 1 , wherein computing the randomized parameter values comprises computing randomized rotation angles, and the method comprises assigning the randomized rotation angles to respective quantum logic gates in the quantum logic circuit.

5. The method of claim 1 , wherein computing the randomized parameter values comprises computing randomized rotation angles, and the method comprises assigning the randomized rotation angles to respective single-qubit rotation gates in the quantum logic circuit.

6. The method of claim 1 , wherein the quantum processor output data represents an output distribution generated by the quantum logic circuit.

7. The method of claim 1 , wherein the quantum logic circuit comprises qubit measurements, and the quantum processor output data comprises one or more bitstrings representing the qubit measurements.

8. The method of claim 7 , wherein the one or more bitstrings representing the qubit measurements comprise a first bitstring, and the data comprises a second, shorter bitstring comprising a subset of bits produced from the first bitstring.

9. The method of claim 1 , comprising executing an iterative process to produce the-data from the input data, wherein the data is derived based on quantum processor output data from multiple episodes of the iterative process, wherein one or more episodes of the iterative process comprises:

obtaining a new set of random numbers;

computing a new set of randomized parameter values based on the input data and the new set of random numbers; and

updating the data based on quantum processor output data produced by an execution of the quantum logic circuit having the new set of randomized parameter values assigned to the variable parameters.

10. The method of claim 9 , wherein the one or more episodes produce an output corresponding to an expectation value of an observable.

11. The method of claim 1 , wherein the input data comprises an input vector, and computing the randomized parameter values comprises:

computing random embedding vectors comprising a first set of random numbers obtained by sampling a first probability distribution;

computing a bias vector comprising a second set of random numbers obtained by sampling a second probability distribution; and

generating a parameter vector by combining the input vector, the random embedding vectors and the bias vector, wherein the parameter vector comprises the randomized parameter values.

12. The method of claim 11 , wherein each of the random embedding vectors acts on a subpart of the input vector to generate the parameter vector.

13. The method of claim 1 , wherein the computer process comprises a machine learning process executable by a classical computer system.

14. The method of claim 1 , wherein the quantum logic circuit having the randomized parameter values assigned to the variable parameters corresponds to an implied kernel function.

15. A hybrid classical-quantum computer comprising:

one or more classical processors; and

memory storing instructions that, when executed by the one or more classical processors, cause the one or more classical processors to perform operations comprising:

obtaining input data for a computer process configured to run on a computer processing unit;

computing randomized parameter values for variable parameters of a quantum logic circuit, wherein the randomized parameter values are computed from the input data and a set of random numbers;

causing one or more quantum processors to generate quantum processor output data, wherein the quantum processor output data are based on an execution of the quantum logic circuit having the randomized parameter values assigned to the variable parameters; and

providing data derived from the quantum processor output data to the computer processing unit configured to run the computer process.

16. The hybrid classical-quantum computer system of claim 15 , the operations comprising obtaining the set of random numbers by sampling a probability distribution.

17. The hybrid classical-quantum computer system of claim 15 , wherein computing the randomized parameter values comprises computing randomized rotation angles, and the method comprises assigning the randomized rotation angles to respective quantum logic gates in the quantum logic circuit.

18. The hybrid classical-quantum computer system of claim 15 , wherein the quantum processor output data represents an output distribution generated by the quantum logic circuit.

19. The hybrid classical-quantum computer system of claim 15 , the operations comprising executing an iterative process to produce the data from the input data, wherein the data is derived based on quantum processor output data from multiple episodes of the iterative process, wherein one or more episodes of the iterative process comprises:

obtaining a new set of random numbers;

computing a new set of randomized parameter values based on the input data and the new set of random numbers; and

updating the data based on quantum processor output data produced by an execution of the quantum logic circuit having the new set of randomized parameter values assigned to the variable parameters.

20. The hybrid classical-quantum computer system of claim 19 , wherein the one or more episodes produce an output corresponding to an expectation value of an observable.

21. The hybrid classical-quantum computer system of claim 15 , wherein the input data comprises an input vector, and computing the randomized parameter values comprises:

computing random embedding vectors comprising a first set of random numbers obtained by sampling a first probability distribution;

computing a bias vector comprising a second set of random numbers obtained by sampling a second probability distribution; and

generating a parameter vector by combining the input vector, the random embedding vectors and the bias vector, wherein the parameter vector comprises the randomized parameter values.

22. The hybrid classical-quantum computer system of claim 15 ,

further comprising the one or more quantum processor units and the computer processing unit, the computer processing unit comprising one or more classical processors configured to:

receive the data as input for the computer process; and

execute the computer process based on the data.