Automated synthesizing and compilation of quantum programs
In a general aspect, a quantum program can be automatically synthesized or compiled. In some implementations, a discretized state space is obtained. The discretized state space includes a plurality of states for one or more qubits of a quantum processor. A discrete action space is obtained. The discrete action space includes a plurality of unitary operations for the one or more qubits of the quantum processor. A policy that uses the discretized state space and the discrete action space to generate quantum programs for quantum state preparation is defined. A dynamic programming process is used to improve the policy. An initial state and a target state of the one or more qubits is identified. The policy is used to generate a quantum program to produce the target state from the initial state. The quantum program include a subset of the unitary operations in the discrete action space.
1. A method comprising:
obtaining, by one or more processors, a discretized state space, the discretized state space comprising a first plurality of unitary operations for one or more qubits of a quantum processor, wherein the first plurality of unitary operations are expressed in a high-level quantum programming language, and the quantum processor permits only a specified set of quantum logic gates;
obtaining, by one or more processors, a discrete action space, the discrete action space comprising a second plurality of unitary operations for the one or more qubits of the quantum processor, wherein the second plurality of unitary operations are expressed in a low-level quantum programming language and correspond to the specified set of quantum logic gates that are permitted by the quantum processor;
defining, by one or more processors, a neural network that uses the discretized state space and the discrete action space to compile quantum programs;
training, by one or more processors, the neural network using a dynamic programming process;
identifying, by one or more processors, a target unitary operation in the discretized state space;
generating, by one or more processors, a quantum program using the trained neural network, the quantum program being configured to perform the target unitary operation using a subset of the second plurality of unitary operations in the discrete action space; and
causing the quantum processor to execute the quantum program.
2. The method of claim 1 , wherein obtaining the discretized state space comprises:
determining a quaternion of each of the first plurality of unitary operations of the discretized state space; and
discretizing the quaternion.
3. The method of claim 2 , wherein discretizing the quaternions comprises:
discretizing four elements of the quaternion, wherein the four elements equal n i ·Δ, n i is an integer, i is a positive integer, and i=1, 2, 3, 4, 4 equals √{square root over (2)}k, k is a number of unitary operations in the quantum program, and is a positive integer.
4. The method of claim 1 , comprising:
for each unitary operation in the discrete action space, computing transition probabilities between unitary operations in the first plurality of unitary operations in the discretized state space; and
using the transition probabilities in the dynamic programming process to train the neural network.
5. The method of claim 1 , wherein training the neural network using the dynamic programming process comprises evaluating a reward based on a Euclidean distance, wherein the reward is configured to train the neural network to use one or more of the unitary operations in the second plurality of unitary operations in the discrete action space to perform the target unitary operation in the discretized state space.
6. The method of claim 5 , wherein the Euclidean distance is defined as an absolute difference between a target quaternion and an evolved quaternion, the target quaternion corresponds to the target unitary operation, and the evolved quaternion is prepared via one or more unitary operations in the second plurality of unitary operations.
7. The method of claim 1 , wherein obtaining the discrete action space comprises discretizing a single-qubit action space.
8. The method of claim 1 , comprising:
generating a plurality of quantum programs using the trained neural network to produce evolved quaternions;
selecting, from the plurality of quantum programs, the quantum program having a shortest length; and
using the selected quantum program to perform the target unitary operation, wherein the evolved quaternions are offset from a target quaternion by respective Euclidean distances, the Euclidean distances being less than a predetermined maximum distance.
9. The method of claim 1 , comprising:
generating a plurality of optimal quantum programs to perform respective target unitary operations;
forming gate compilation sequences by chaining the plurality of optimal quantum programs; and
providing the gate compilation sequences for execution on the quantum processor.
10. The method of claim 1 , wherein:
the quantum processor comprises a superconducting circuit comprising a plurality of qubit devices each comprising one or more Josephson junctions;
the second plurality of unitary operations correspond to quantum logic gates that are executable by specific hardware of the quantum processor; and
the method comprises executing the quantum program using the specific hardware of the quantum processor, wherein executing the quantum program comprises delivering control signals to the superconducting circuit to perform one or more of the second plurality of unitary operations.
11. A computing system comprising:
one or more processors; and
memory storing instructions that are configured, when executed by the one or more processors, to perform operations comprising:
obtaining a discretized state space, the discretized state space comprising a first plurality of unitary operations for one or more qubits of a quantum processor, wherein the first plurality of unitary operations is expressed in a high-level quantum programing language, and the quantum processor permits only a specified set of quantum logic gates;
obtaining a discrete action space, the discrete action space comprising a second plurality of unitary operations for the one or more qubits of the quantum processor, wherein the second plurality of unitary operations are expressed in a low-level quantum programing language and correspond to the specified set of quantum logic gates that are permitted by the quantum processor;
defining a neural network that uses the discretized state space and the discrete action space to compile quantum programs;
training the neural network using a dynamic programming process;
identifying a target unitary operation in the discretized state space;
generating a quantum program using the trained neural network, the quantum program being configured to perform the target unitary operation using a subset of the second plurality of unitary operations in the discrete action space; and
causing the quantum processor to execute the quantum program.