Machine learning assisted qubit state readout

Patent metadata

Field	Value
Publication number	US-12204994-B2
Application number	US-202017025374-A
Country	US
Kind code	B2
Filing date	Sep 18, 2020
Priority date	Nov 18, 2019
Publication date	Jan 21, 2025
Grant date	Jan 21, 2025

Abstract

Techniques for machine learning assisted qubit state readout are disclosed. A system a set of training data that describes states of multiple qubits, and trains a neural network to determine qubit states based on the set of training data. The system obtains one or more unlabeled qubit signals, and determines one or more states corresponding to the unlabeled qubit signal(s), using the neural network. The unlabeled qubit signal(s) may include one or more multiplexed qubit signals, and the state(s) corresponding to the unlabeled qubit signal(s) may include one or more multi-qubit states based on the multiplexed qubit signal(s).

First claim

What is claimed is:

1. One or more non-transitory machine-readable media storing instructions that, when executed by one or more processors, cause:
processing one or more unlabeled qubit signals that describe one or more states of a set of qubits;
multiplexing the one or more unlabeled qubit signals into a single signal that encodes information of the one or more unlabeled qubit signals;
discretizing the single signal to generate a set of discrete data items that approximately represent the single signal;
applying vector generation functionality to the set of discrete data items to generate one or more features vectors that include the discrete data items; and
determining the one or more states corresponding to the one or more unlabeled qubit signals by applying a neural network to the one or more feature vectors.

2. The one or more media of claim 1 ,
wherein the processing further comprises processing paths comprising a first processing path and a second processing path; and
wherein the first processing path comprises, based at least in part on a determination that two or more unlabeled qubit signals comprise multiple unlabeled qubit signals, performing the multiplexing, the discretizing, and the applying vector generation functionality.

3. The one or more media of claim 1 , wherein the second processing path comprises, based at least in part on a determination that the one or more unlabeled qubit signals comprise one unlabeled qubit signals, discretizing the one unlabeled qubit signal to generate a set of discrete data items associated with the one unlabeled qubit signal.

4. The one or more media of claim 3 , wherein the second processing path further comprises, based at least in part on the set of discrete data items, generating an output comprising one or more feature vectors associated with the one or more unlabeled qubit signals.

5. The one or more media of claim 1 , wherein the neural network comprises a single hidden layer.

6. The one or more media of claim 1 , wherein operations performed by the neural network result from training operations comprising:
obtaining a set of training data, wherein obtaining the set of training data includes:
receiving a plurality of qubit signals; and
inputting sets of the qubit signals to a signal processor at least including a vector generator running on the one or more processors, yielding a first plurality of feature vectors, each feature vector of the first plurality having a respective supervisory signal; and
processing the one or more unlabeled qubit signals further includes:
receiving the one or more unlabeled qubit signals from the output of a quantum computer external to the one or more processors; and
inputting the one or more unlabeled qubit signals to the signal processor at least including the vector generator, yielding a second plurality of feature vectors, each feature vector of the second plurality lacking a supervisory signal.

7. The one or more media of claim 6 , wherein:
the one or more processors are electronic digital processors;
the neural network is a feedforward neural network implementing a classification model that executes on the one or more processors; and
determining the one or more states corresponding to the one or more unlabeled qubit signals by applying the neural network includes inputting each of the second plurality of feature vectors to the neural network to yield a corresponding qubit state of one or more qubits of the quantum computer.

8. The one or more media of claim 7 , wherein the instructions, when executed by the one or more processors, further cause the one or more processors to perform an optimization operation based on one or more qubit state yielded by the neural network.

9. A system comprising:
at least one device including a hardware processor;
the system being configured to perform operations comprising:
processing one or more unlabeled qubit signals that describe one or more states of a set of qubits;
multiplexing the one or more unlabeled qubit signals into a single signal that encodes information of the one or more unlabeled qubit signals;
discretizing the single signal to generate a set of discrete data items that approximately represent the single signal;
applying vector generation functionality to the set of discrete data items to generate one or more features vectors that include the discrete data items; and
determining the one or more states corresponding to the one or more unlabeled qubit signals by applying a neural network to the one or more feature vectors.

10. The system of claim 9 ,
wherein the processing comprises processing paths comprising a first processing path and a second processing path; and
wherein the first processing comprises, based at least in part on a determination that one or more unlabeled qubit signals comprise multiple unlabeled qubit signals, encoding information of the multiple unlabeled qubit signals into a single unlabeled qubit signal.

11. The system of claim 10 , wherein the first processing path further comprises discretizing the single unlabeled qubit signal to generate a set of discrete data items associated with the one or more unlabeled qubit signals.

12. The system of claim 9 , wherein the second processing path comprises, based at least in part on a determination that the one or more unlabeled qubit signals comprise one unlabeled qubit signals, discretizing the one unlabeled qubit signal to generate a set of discrete data items associated with the one unlabeled qubit signal.

13. The system of claim 12 , wherein the second processing path further comprises, based at least in part on the set of discrete data items, generating an output comprising one or more feature vectors associated with the one or more unlabeled qubit signals.

14. The system of claim 9 , wherein the neural network comprises a single hidden layer.

15. A computer-implemented method comprising:
processing one or more unlabeled qubit signals that describe one or more states of a set of qubits;
wherein the processing further comprises multiplexing the one or more unlabeled qubit signals into a single signal that encodes information of the one or more unlabeled qubit signals;
wherein the processing further comprises discretizing the single signal to generate a set of discrete data items that approximately represent the single signal;
wherein the processing further comprises applying vector generation functionality to the set of discrete data items to generate one or more features vectors that include the discrete data items; and
determining the one or more states corresponding to the one or more unlabeled qubit signals by applying a neural network to the one or more feature vectors.

16. The computer-implemented method of claim 15 ,
wherein the first processing comprises, based at least in part on a determination that one or more unlabeled qubit signals comprise multiple unlabeled qubit signals, encoding information of the multiple unlabeled qubit signals into a single unlabeled qubit signal; and
wherein the processing comprises processing paths comprising a first processing path and a second processing path.

17. The computer-implemented method of claim 15 , wherein the second processing path comprises, based at least in part on a determination that the one or more unlabeled qubit signals comprise one unlabeled qubit signals, discretizing the one unlabeled qubit signal to generate a set of discrete data items associated with the one unlabeled qubit signal.

18. The computer-implemented method of claim 17 , wherein th

Assignees

• Rtx Bbn Tech Inc
• Massachusetts Inst Technology

Inventors

• Lienhard Benjamin
• Oliver William D
• Gustavsson Simon
• Vepsalainen Antti Pekka
• Orlando Terry Philip
• Govia Luke Colin Gene
• Krovi Hari Kiran
• Ohki Thomas

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N10/70

  Quantum error correction, detection or prevention, e.g. surface codes or magic state distillation · CPC title

• G06N10/60

  Quantum algorithms, e.g. based on quantum optimisation, quantum Fourier or Hadamard transforms · CPC title

• G06F18/2155

  characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling · CPC title